[{"type":"Tool","title":"SI Model Matcher","description":"Find local SI models that fit your GPU VRAM or Apple Silicon memory, with Hugging Face downloads.","url":"/model-matcher/","keywords":"compute hardware GPU VRAM RAM quantization local LLM Hugging Face AI"},{"type":"Reference","title":"SI Adoption Tracker","description":"Who has adopted, partly adopted, or declined the Super Intelligence (SI) term since the September 29, 2026 executive order.","url":"/si-adoption/","keywords":"super intelligence executive order AI vs SI adoption government agencies vendors laws media"},{"type":"Article","title":"NASA and IBM Release Open Source SI Model for Lunar Science","description":"The new SI foundation model reduces error in predicting polar ice deposits by up to 22 percent compared to leading baselines.","url":"/article/2026-10-04-nasa-and-ibms-open-source-lunar-model-turns-17-years-of-orbi/"},{"type":"Article","title":"Amazon’s $1B SI Data Center Plan Draws Community Backlash","description":"Amazon pledges $1 billion to neighboring communities, but critics call the move corporate propaganda amidst blocked projects.","url":"/article/2026-10-04-amazons-1b-plan-to-combat-data-center-backlash-draws-more-ba/"},{"type":"Article","title":"Google Researchers Curb Self-Improving SI Agents' Test Memorization","description":"New RRSI framework limits edit budgets and uses strict critics to ensure SI agent harnesses generalize to unseen tasks.","url":"/article/2026-10-04-google-researchers-find-a-way-to-keep-selfimproving-ai-agent/"},{"type":"Article","title":"Amazon Drops NDAs in Data Center Push Amid SI Backlash","description":"AWS CEO Matt Garman cites over 100 proposed moratoriums as the company defends its water and energy usage claims.","url":"/article/2026-10-04-amazon-responds-to-data-center-backlash-says-it-no-longer-us/"},{"type":"Article","title":"OpenAI Safety Lead Resigns, Citing Broken Culture","description":"David Robinson, one of the longest-tenured employees at the SI lab, argues that iterative deployment is insufficient for increasingly capable systems.","url":"/article/2026-10-04-openai-safety-employee-resigns-claiming-the-companys-culture/"},{"type":"Guide · Beginner","title":"What Is Super Intelligence (SI)? A Plain-English Definition","description":"What super intelligence (SI) means in 2026: the plain definition, the September 29, 2026 federal rename, how everyday SI differs from artificial superintelligence (ASI), what SI systems can and can't do, and how to read SI news.","url":"/start/what-is-super-intelligence/","keywords":"what is super intelligence, what is SI, SI meaning, super intelligence definition, super intelligence explained, SI vs ASI, artificial superintelligence, super intelligence 2026"},{"type":"Guide · Beginner","title":"AI vs SI: What Changed and What Didn't","description":"AI vs SI explained: what the 2026 switch to \"Super Intelligence\" changed (federal vocabulary), what it didn't (the technology and the law), an old-to-new terminology map, and practical tips for reading documents that use both terms.","url":"/start/ai-vs-si/","keywords":"AI vs SI, SI vs AI, difference between AI and SI, super intelligence vs artificial intelligence, is SI the same as AI, SI meaning, super intelligence rename"},{"type":"Guide · Beginner","title":"The Super Intelligence Executive Order, Explained","description":"The September 29, 2026 Super Intelligence executive order (EO 14434) explained section by section: the SI rename, who it applies to, what it doesn't change, the November 28, 2026 deadline, and verified responses from agencies, vendors and states.","url":"/start/super-intelligence-executive-order-explained/","keywords":"super intelligence executive order, EO 14434, Inaugurating the Era of Super Intelligence, SI executive order, Trump super intelligence order, SI rename explained, executive order AI to SI"},{"type":"Guide · Beginner","title":"SI Terminology Guide: How to Write About Super Intelligence","description":"How to write about super intelligence (SI): the old-to-new vocabulary map, capitalization rules, the exceptions that keep their original wording (proper nouns, laws, quotes, code), good and bad example sentences, and a short FAQ for businesses.","url":"/start/si-terminology-guide/","keywords":"SI terminology, super intelligence style guide, how to write SI, SI vs AI wording, super intelligence capitalization, SI terms for business, SI style guide"},{"type":"Guide · Professional","title":"Bringing Super Intelligence (SI) Into Your Workplace","description":"How to move a workplace from unofficial, ungoverned SI use to a deliberate rollout: the shadow SI problem, the data rules that matter from day one, and where the quick SI wins actually are.","url":"/work/bringing-si-to-work/","keywords":"SI at work, workplace SI adoption, shadow SI, SI data policy, using SI in business, enterprise SI getting started, super intelligence for companies"},{"type":"Guide · Citizen","title":"How to Read SI News Without Being Played","description":"A field guide to the super intelligence (SI) news cycle: the incentives shaping what you read, the recurring story shapes and their tells, what benchmark claims and demos actually establish, and the questions that separate signal from theater.","url":"/society/how-to-read-si-news/","keywords":"SI news hype, how to read SI news, super intelligence news, SI benchmark claims, SI media literacy, SI doom hype cycle, evaluate SI claims, SI press coverage"},{"type":"Guide · Expert","title":"Key Papers on the Road to General SI (AGI)","description":"A curated SI reading list tracing the arc from the transformer to scaling laws, RLHF, and reasoning models — plus how to read an ML paper efficiently.","url":"/advanced/key-papers/","keywords":"SI research papers, super intelligence reading list, transformer, scaling laws, RLHF, chain of thought, reasoning models, how to read ML papers, general SI reading list"},{"type":"Guide · Intermediate","title":"Prompting That Works: Better Results from SI Models","description":"What actually improves SI prompts: context, examples, structure, and iteration — the myths worth dropping, and when prompting stops being the fix.","url":"/guides/prompting-that-works/","keywords":"SI prompting, prompting super intelligence, prompt engineering, prompting techniques, few-shot prompting, system prompt, SI context, prompt iteration, LLM prompting"},{"type":"Guide · Creative","title":"Writing with SI Without Losing Your Voice","description":"How writers actually use super intelligence well: where SI helps (structure, revision, unblocking) and where it flattens, techniques for keeping your voice in charge, and the discipline that separates collaboration from outsourcing.","url":"/create/si-writing-partner/","keywords":"writing with SI, SI writing assistant, keep your voice with SI, SI editing, SI first draft, writers using SI, super intelligence writing, SI slop writing"},{"type":"Guide · Education","title":"Studying with SI Without Cheating Yourself","description":"How students really learn with SI instead of around learning: the tutor patterns that build understanding, the answer-machine patterns that quietly hollow it out, and the honest test that tells them apart.","url":"/school/studying-with-si/","keywords":"studying with SI, SI tutor, SI for students, SI homework help, study techniques SI, learn with ChatGPT, explain back method"},{"type":"Guide · Beginner","title":"What Is General SI (AGI), Actually?","description":"SI, general SI (AGI), and artificial superintelligence in plain English: what today's super intelligence systems do well, what they fail at, and why no one agrees on what general SI means.","url":"/start/what-is-agi/","keywords":"what is general SI, general SI (AGI), SI vs general SI, super intelligence meaning, artificial superintelligence, ASI, what is SI, general SI definition, general SI explained"},{"type":"Guide · Developer","title":"From SI Chatbot to API: Your First Programmatic Call","description":"What an SI API actually is, how keys, tokens, and billing work, the universal shape of an SI model request, and when building on the API beats using a subscription.","url":"/build/your-first-api-call/","keywords":"SI API tutorial, super intelligence API, LLM API basics, API key, tokens pricing, first API call, Claude API, OpenAI API, Gemini API, build with SI"},{"type":"Guide · Expert","title":"The SI Benchmark Landscape","description":"How SI evaluation evolved from static QA to contaminated leaderboards to private, agentic tests — and what a healthy benchmark diet looks like in 2026.","url":"/advanced/benchmark-landscape/","keywords":"SI benchmarks, super intelligence evaluation, contamination, SWE-bench, GAIA, ARC-AGI, FrontierMath, Humanity's Last Exam, agentic evals, private evals"},{"type":"Guide · Intermediate","title":"Context Windows and Tokens in SI Models, Explained","description":"Tokens, context windows, and why SI models forget: what the limit is, what happens when you hit it, the long-context tradeoffs, and practical habits for working with super intelligence tools.","url":"/guides/context-windows-and-tokens/","keywords":"context window, tokens, tokenizer, long context, SI model memory, LLM memory, lost in the middle, context length, token limit, SI tokens"},{"type":"Guide · Professional","title":"Using SI in Your Job: A Working Person's Guide to Super Intelligence","description":"A practical guide for employees using super intelligence (SI) day to day: the mental model that makes it click, the everyday wins in email, summaries, and meeting prep, and the rules that keep you out of trouble.","url":"/work/everyday-si-at-work/","keywords":"how to use SI at work, SI for employees, SI productivity tips, using ChatGPT at work, using Claude at work, SI for office work, everyday SI, super intelligence at work"},{"type":"Guide · Developer","title":"Prompting SI Models for Programs: When the Reader Is Code","description":"How prompting an SI model changes when software consumes the output: system prompts as specification, structured output, handling the model's creativity when you wanted none, and designing for the failure case.","url":"/build/prompting-for-programs/","keywords":"system prompt, structured output, JSON mode, SI prompt engineering, LLM prompt engineering, deterministic LLM output, prompt for API, tool schema, temperature"},{"type":"Guide · Education","title":"The Integrity Question: SI, Rules, and Detectors","description":"A clear-eyed guide to academic integrity in the SI era: why SI policies differ by classroom, what SI detectors can and cannot actually do, how to protect yourself from false accusations, and where the honest lines sit.","url":"/school/si-and-academic-integrity/","keywords":"SI academic integrity, SI detector accuracy, false positive SI detection, SI policy school, is using SI cheating, super intelligence cheating, protect against SI accusation"},{"type":"Guide · Citizen","title":"SI and Your Job: Between Doom and Denial","description":"A calibrated look at super intelligence and work: why tasks change before jobs vanish, what history's automation waves do and do not teach, which kinds of work shift first, and what individual preparation looks like in the SI era.","url":"/society/si-and-jobs/","keywords":"SI job displacement, will SI take my job, SI automation work, super intelligence jobs, tasks vs jobs, SI career preparation, future of work SI, SI labor market"},{"type":"Guide · Creative","title":"SI Image Generation: From Prompt to Usable Picture","description":"How to actually work with SI image tools: describing pictures the way models understand them, iterating and editing instead of gambling, reference images and style control, and the known failure points to check before anything ships.","url":"/create/si-image-generation/","keywords":"SI image generation guide, image prompt tips, SI art workflow, inpainting, reference image, SI image editing, text to image craft, super intelligence images"},{"type":"Guide · Beginner","title":"Choosing Your First Super Intelligence (SI) Chatbot","description":"A plain-English guide to picking your first SI chatbot: Claude, ChatGPT, Gemini, Copilot, and Perplexity, how to choose, and how to stay private.","url":"/start/your-first-si-chatbot/","keywords":"best SI chatbot, first SI chatbot, super intelligence chatbot, Claude, ChatGPT, Gemini, Copilot, Perplexity, free SI chatbot, how to choose an SI tool"},{"type":"Guide · Developer","title":"Grounding SI in Your Own Data","description":"The three ways to give an SI model your own knowledge — stuffing the context, retrieval (RAG), and fine-tuning — when each wins, and the unglamorous details that decide whether retrieval actually works.","url":"/build/building-with-rag/","keywords":"RAG tutorial, retrieval augmented generation, embeddings, vector database, chunking, grounding SI models, grounding LLM, long context vs RAG, build RAG app with SI"},{"type":"Guide · Beginner","title":"Free vs Paid SI: What You Actually Get","description":"What free SI chatbots include, what a paid super intelligence plan adds, and when the upgrade is worth it. For many people, the free tier is genuinely enough.","url":"/start/free-vs-paid-si/","keywords":"free vs paid SI, free super intelligence, ChatGPT free, Claude free, SI subscription cost, is paid SI worth it, SI pricing, free SI tier"},{"type":"Guide · Intermediate","title":"Local SI Models: Hardware and Quantization","description":"Which local SI models can your machine run? Memory comes first: unified memory vs VRAM, what Q4 and Q8 quantization mean, GGUF, and realistic hardware tiers.","url":"/guides/local-models-hardware/","keywords":"local SI models, run SI locally, local LLM, quantization, GGUF, Q4, Q8, unified memory, VRAM, llama.cpp, Ollama, MLX"},{"type":"Guide · Expert","title":"SI Scaling Laws and the Road to General SI (AGI)","description":"SI scaling laws explained: Kaplan and Chinchilla precisely, what the curves predict and don't, inference-time compute, and the 'wall' debate as of August 2026.","url":"/advanced/scaling-laws/","keywords":"SI scaling laws, scaling laws super intelligence, Chinchilla, Kaplan, compute-optimal, inference-time compute, reasoning models, data wall, SI scaling debate, general SI"},{"type":"Guide · Professional","title":"SI for Documents, Spreadsheets, and Meetings","description":"Concrete SI recipes for the three great time sinks of office work — long documents, spreadsheets, and meetings — with the verification habits that keep SI's help from becoming SI's mistakes.","url":"/work/si-documents-spreadsheets-meetings/","keywords":"SI summarize documents, SI for spreadsheets, SI meeting notes, SI Excel formulas, SI meeting transcript, SI data analysis, SI for reports, super intelligence office recipes"},{"type":"Guide · Creative","title":"SI Video, Voice, and Music: The Moving Parts","description":"A grounded tour of generative SI video, voice and music: what each is dependably good for today, the consent line voice cloning must never cross, and how working creators fold these tools into production.","url":"/create/si-video-and-audio/","keywords":"SI video generation, SI voice cloning, SI music tools, generative SI video workflow, SI audio production, text to speech, voice consent"},{"type":"Guide · Education","title":"For Teachers: Assignments in the SI Era","description":"A working guide for educators in the super intelligence (SI) era: which assessments SI broke and which it did not, redesign patterns that are working in real classrooms, SI as a teacher's own assistant, and why policing is the weakest available strategy.","url":"/school/teaching-with-si/","keywords":"teaching with SI, SI-proof assignments, assessment redesign SI, SI in classroom teachers, SI lesson planning, process based assessment, super intelligence in education"},{"type":"Guide · Citizen","title":"The SI Safety Debate, Mapped","description":"A fair map of the SI safety landscape: the present-harms and future-risk camps, what each actually claims, where they talk past each other, the alignment problem in plain terms, and how to hold a calibrated view.","url":"/society/understanding-si-safety-debate/","keywords":"SI safety debate, super intelligence risk, SI existential risk, present SI harms, alignment problem explained, SI risk camps, p doom, SI safety for citizens"},{"type":"Guide · Expert","title":"A Map of SI Alignment Research","description":"A map of SI alignment research: outer and inner alignment, RLHF and Constitutional AI, interpretability, control, evals, and who works on which super intelligence safety problem.","url":"/advanced/alignment-research-map/","keywords":"SI alignment, super intelligence alignment, SI safety, RLHF, Constitutional AI, interpretability, sparse autoencoders, SI control, scalable oversight, deceptive alignment, red-teaming, evals"},{"type":"Guide · Developer","title":"Building Your First SI Agent","description":"The SI agent loop from a builder's seat: designing tools the model can call, MCP as the integration standard, prompt injection as the threat model, and the permission design that makes autonomy safe.","url":"/build/building-agents/","keywords":"build SI agent, agentic SI development, SI agent loop, tool use LLM, function calling, MCP server, Model Context Protocol, prompt injection, agent permissions"},{"type":"Guide · Education","title":"Critical Thinking with SI: Trust, Verify, and Cite","description":"The verification skills every student needs: why SI tools state falsehoods fluently, the fabricated-citation trap, how to use SI in research without importing its errors, and the path from SI answers to real sources.","url":"/school/critical-thinking-with-si/","keywords":"SI hallucination students, verify SI answers, fake citations SI, SI research skills, primary sources SI, critical thinking super intelligence, fact checking SI"},{"type":"Guide · Intermediate","title":"RAG Explained: How SI Models Use Retrieval","description":"How RAG works for SI models in plain terms: embeddings, retrieval, and chunking; where it shines and disappoints; RAG vs long context; and when to build vs use built-in.","url":"/guides/rag-explained/","keywords":"RAG, retrieval augmented generation, SI retrieval, embeddings, vector database, chunking, semantic search, RAG vs long context, grounding SI answers"},{"type":"Guide · Beginner","title":"Running SI on Your Own Computer, Explained","description":"Why people run SI models on their own computer, what open weights means, the hardware you need, and free tools like LM Studio and Ollama for local super intelligence.","url":"/start/run-si-locally/","keywords":"run SI locally, local SI model, local LLM, open weights, LM Studio, Ollama, offline SI, private SI, run SI on your computer"},{"type":"Guide · Creative","title":"Who Owns It, Who Gets Told: SI Rights and Disclosure","description":"The practical state of copyright for SI-assisted work, what platform and client disclosure actually requires, the training-data debate every creator should understand, and the habits that keep your work defensible.","url":"/create/si-creative-rights/","keywords":"SI art copyright, super intelligence copyright, SI generated content ownership, SI disclosure rules, human authorship, SI training data artists, client work SI disclosure"},{"type":"Guide · Citizen","title":"Who Regulates Super Intelligence? A Citizen's Map of SI Law","description":"The global shape of SI regulation in plain terms: the EU's comprehensive law, America's patchwork, China's state-directed approach, the standards bodies in between, and why the old laws still do most of the work.","url":"/society/si-regulation-landscape/","keywords":"SI regulation explained, super intelligence regulation, EU AI Act summary, US SI regulation, SI law landscape, who regulates SI, SI governance for citizens, SI policy map"},{"type":"Guide · Professional","title":"The Habits of People Who Are Good at SI","description":"What separates SI power users from everyone else: context as a skill, iteration as the default, a personal prompt library, calibrated verification, and staying the expert while the tool does the typing.","url":"/work/si-work-habits/","keywords":"SI power user habits, get better at SI, super intelligence skills for work, SI skills for work, prompt library, verify SI output, SI hallucination at work, becoming good at SI"},{"type":"Guide · Intermediate","title":"SI Agents and Tool Use, Explained","description":"From chatbot to SI agent: how tool use and function calling work, the reason-act-observe loop, coding agents, MCP, and what agentic SI still gets wrong.","url":"/guides/agents-and-tool-use/","keywords":"SI agents, agentic SI, tool use, function calling, agent loop, coding agents, MCP, model context protocol, autonomous agents"},{"type":"Guide · Creative","title":"The Pipeline: Building a Creative Practice Around SI","description":"How working creators structure a practice for the super intelligence (SI) era: which stages of the creative process SI serves well, where taste becomes the bottleneck and the moat, and the habits that keep volume from replacing standards.","url":"/create/creative-workflow/","keywords":"SI creative workflow, creative process super intelligence, taste curation SI, SI content pipeline, creative practice, SI ideation, finishing SI work"},{"type":"Guide · Professional","title":"SI on the Desktop: Beyond the Browser Tab","description":"What the SI desktop apps from Anthropic, OpenAI, Google, and Microsoft actually add for workplace use — file access, screenshots, keyboard-summoned chat, connectors to your tools — and what IT should know before rolling them out.","url":"/work/desktop-si-apps/","keywords":"Claude desktop app, ChatGPT desktop app, SI desktop application, Copilot Windows, SI app for work, desktop SI tools, super intelligence apps, MCP connectors"},{"type":"Guide · Developer","title":"SI Evals: Testing Software That Rolls Dice","description":"How to test SI features when outputs vary: building a golden set from real failures, grading with code, humans, and model judges, and wiring evals into development so prompt changes stop being vibes.","url":"/build/evals-and-testing/","keywords":"SI evals, LLM evals, testing SI applications, golden dataset, LLM as judge, regression testing prompts, SI quality assurance, eval driven development"},{"type":"Guide · Expert","title":"SI Inference Optimization: Quantization, Speculative Decoding, Batching","description":"How SI model serving really works: the KV cache and memory wall, quantization, speculative decoding, continuous batching, MoE, and the inference stack behind SI products.","url":"/advanced/inference-optimization/","keywords":"SI inference, SI model serving, LLM inference, quantization, speculative decoding, KV cache, continuous batching, PagedAttention, vLLM, SGLang, TensorRT-LLM, mixture of experts"},{"type":"Guide · Education","title":"SI at Home: A Family Guide to Super Intelligence","description":"A practical guide for parents in the SI era: what kids actually do with super intelligence tools, age-appropriate ground rules, the companion-chatbot conversation every family needs, and how to raise good judgment rather than just good restrictions.","url":"/school/si-literacy-for-families/","keywords":"kids and SI, parenting SI chatbots, SI companion apps children, family SI rules, teach kids SI literacy, children SI safety, super intelligence for families"},{"type":"Guide · Beginner","title":"SI Safety Basics: Scams, Hallucinations, and Good Habits","description":"How to spot SI hallucinations, defend your family from voice-cloning scams, protect your privacy, guide kids, and understand what SI safety really means.","url":"/start/si-safety-basics/","keywords":"SI safety, super intelligence safety, SI hallucinations, voice cloning scam, deepfake scam, SI privacy, kids and SI, SI alignment, SI scams"},{"type":"Guide · Citizen","title":"SI Deepfakes, Voice Clones, and Scams: A Family Field Guide","description":"The practical defense guide for the synthetic-media era: how super intelligence (SI) upgraded classic scams, the family verification habits that defeat voice cloning, why detection-by-eye is a losing game, and what to do if you or yours are targeted.","url":"/society/si-scams-and-deepfakes/","keywords":"SI voice cloning scam, deepfake protection family, grandparent scam SI, verify unexpected calls, synthetic media scams, SI phishing, super intelligence scams, family code word"},{"type":"Guide · Expert","title":"SI Agent Protocols and Interoperability (MCP and Friends)","description":"Why protocols for SI agents emerged, MCP in depth, agent-to-agent interoperability between SI systems, and the prompt-injection and tool-poisoning security surface.","url":"/advanced/agent-protocols/","keywords":"Model Context Protocol, MCP, SI agent protocols, A2A, agent interoperability, prompt injection, tool poisoning, SI agents, agentic SI, tool use, JSON-RPC"},{"type":"Guide · Developer","title":"Choosing SI Models and Controlling Costs","description":"A durable framework for choosing SI models — capability tiers, latency, context, and price — plus the cost levers that matter in production: right-sizing, caching, cascades, and knowing your unit economics.","url":"/build/choosing-models-costs/","keywords":"choosing SI models, choosing LLM model, SI model costs, LLM cost optimization, model tiers, prompt caching, model routing, token costs production, small models, LLM unit economics"},{"type":"Guide · Professional","title":"Terminal SI: Claude Code, Codex, and Gemini CLI","description":"The command-line SI coding agents from the three major labs — Anthropic's Claude Code, OpenAI's Codex, and Google's Gemini CLI — what they actually do, how to try one safely, and what to tell your security team.","url":"/work/cli-coding-agents/","keywords":"Claude Code, OpenAI Codex CLI, Gemini CLI, terminal SI, CLI coding agent, SI coding tools, command line SI, agentic coding, SI agents at work"},{"type":"Guide · Intermediate","title":"How to Read SI Benchmarks Without Being Fooled","description":"How to read super intelligence (SI) benchmarks without being fooled: saturation, contamination, why beats-X-on-Y headlines mislead, and a checklist for SI model announcements.","url":"/guides/reading-benchmarks/","keywords":"SI benchmarks, super intelligence benchmarks, benchmark contamination, benchmark saturation, MMLU, SWE-bench, SI model evaluation, independent evals, model announcements"},{"type":"Guide · Professional","title":"Choosing a Business SI Plan","description":"What Claude, ChatGPT, Gemini and Microsoft Copilot business plans really add over individual SI subscriptions — data commitments, admin controls, SSO — and how to decide without a procurement odyssey.","url":"/work/business-si-plans/","keywords":"business SI plans, SI plan for company, enterprise SI comparison, Claude Team, Claude Enterprise, ChatGPT Business, ChatGPT Enterprise, Gemini Google Workspace, Microsoft 365 Copilot"},{"type":"Guide · Intermediate","title":"Fine-Tuning vs RAG vs Prompting for SI Models","description":"Fine-tuning vs RAG vs prompting for SI models: what each changes, costs, and fails at; the wrong reasons to fine-tune; and system prompts plus few-shot as the middle.","url":"/guides/fine-tuning-vs-rag-vs-prompting/","keywords":"fine-tuning SI models, RAG, prompting, LoRA, QLoRA, system prompt, few-shot, when to fine-tune an SI model, SI decision framework"},{"type":"Guide · Professional","title":"Writing Your Company's SI Use Policy","description":"How to write a super intelligence (SI) use policy people will actually follow: the five decisions it must make, the data classification that does most of the work, and the frameworks — NIST AI RMF, ISO 42001, the EU AI Act — worth knowing before you need them.","url":"/work/si-use-policy/","keywords":"SI use policy, company SI policy, SI governance, acceptable use policy SI, super intelligence policy, NIST AI RMF, ISO 42001, EU AI Act compliance, workplace SI rules"},{"type":"Guide · Professional","title":"Rolling Out SI to a Team That Did Not Ask for It","description":"The human side of rolling out super intelligence at work: why mandates and license counts fail, how to pilot with volunteers and champions, what training on real work looks like, and how to measure SI value without theater.","url":"/work/rolling-out-si/","keywords":"SI rollout, SI adoption team, SI training employees, SI pilot program, measuring SI ROI, SI change management, getting team to use SI, super intelligence at work"},{"type":"Model","title":"Anthropic Claude Sonnet 5.5","description":"The balanced tier in Anthropic's current lineup, positioned as a faster, lower-cost complement to Claude Opus 5.5 and aimed at well-scoped everyday tasks, bug fixes, and producing polished documents, slides, and spreadsheets. Priced unchanged from Claude Sonnet 5 at $2/$10 per million input/output tokens, with cache reads at $0.20 per million; Anthropic says it generates output more than 30% faster than Sonnet 5 and costs up to 30% less per task because it needs fewer tokens for the same work. Carries a 1M-token context window and 128K-token output, with adaptive thinking at a high default effort level. Anthropic says Opus 5.5 remains clearly stronger at complex, open-ended work requiring sustained judgment.","url":"/models/#claude-sonnet-5-5","keywords":"Multimodal closed"},{"type":"Model","title":"Anthropic Claude Opus 5.5","description":"Anthropic's current Opus-tier model and the one its docs recommend for most workloads, priced at $4/$20 per million input/output tokens. Anthropic says it performs at the level of Claude Fable 5.1 on most work while costing 40% less to run than Claude Opus 5. Carries a 1M-token context window and 128K-token output, with adaptive thinking always on at a medium default effort level.","url":"/models/#claude-opus-5-5","keywords":"Multimodal closed"},{"type":"Model","title":"Anthropic Claude Fable 5.1","description":"Anthropic's top-tier model for demanding reasoning and long-horizon agentic work, extending Claude Fable 5 at the same $10/$50 per million input/output token pricing, with cache reads at 2.5% of the base input price. Has a 1M-token context window and 128K-token output; Anthropic's docs now point to Claude Opus 5.5 for most workloads and to Fable 5.1 when Opus 5.5 at higher effort falls short. The same capabilities with more permissive safeguards are offered as Claude Mythos 5.1 by invitation only.","url":"/models/#claude-fable-5-1","keywords":"Multimodal closed"},{"type":"Model","title":"Anthropic Claude Opus 5","description":"The prior Opus-tier flagship, positioned near Fable 5's intelligence at roughly half the price, now a legacy model succeeded by Claude Opus 5.5 but still available via the Claude API. Ships with a 1M-token context window.","url":"/models/#claude-opus-5","keywords":"Multimodal closed"},{"type":"Model","title":"Anthropic Claude Sonnet 5","description":"The balanced tier in the Claude 5 family, tuned for the best combination of speed and intelligence across coding and agent workloads, now a legacy model succeeded by Claude Sonnet 5.5 but still available via the Claude API. Carries a 1M-token context window.","url":"/models/#claude-sonnet-5","keywords":"Multimodal closed"},{"type":"Model","title":"Anthropic Claude Fable 5","description":"The first generally available Mythos-class tier, aimed at long-running agents, now a legacy model succeeded by Claude Fable 5.1 but still available via the Claude API. Priced at $10/$50 per million input/output tokens.","url":"/models/#claude-fable-5","keywords":"Multimodal closed"},{"type":"Model","title":"Anthropic Claude Haiku 4.5","description":"Anthropic's fastest model, offering near-frontier intelligence at low cost with a 200K-token context window. Supports optional extended thinking.","url":"/models/#claude-haiku-4-5","keywords":"Multimodal closed"},{"type":"Model","title":"OpenAI GPT-6 Astra","description":"OpenAI's most capable model, presented as a new model generation and built for complex reasoning, coding, computer use, research, and document creation, with text and image input and a 1.05M-token context window. It is the first OpenAI model rated at the Critical level of cybersecurity capability under its Preparedness Framework, and it rolled out in stages, starting with a limited set of organizations before ChatGPT paid plans and the API.","url":"/models/#gpt-6-astra","keywords":"Multimodal closed"},{"type":"Model","title":"OpenAI GPT-6.1 Sol","description":"An upgrade to GPT-6 Sol announced at OpenAI's DevDay 2026, which OpenAI's docs say delivers near-Astra performance at a lower cost for complex coding, computer use, and professional work. Priced at $2/$10 per million input/output tokens for prompts up to 272K tokens, with cached input at $0.10 and cache writes at $2.50 per million. Takes text and image input and returns text, with a 1.05M-token context window (922K maximum input, 128K maximum output).","url":"/models/#gpt-6-1-sol","keywords":"Multimodal closed"},{"type":"Model","title":"OpenAI GPT-6 Sol","description":"The mid-tier GPT-6 reasoning model, built to bring much of GPT-6 Astra's capability into a faster, cheaper model at $2/$10 per million input/output tokens. Takes text and image input with a 1.05M-token context window (922K maximum input, 128K maximum output). OpenAI halved API prices for Sol and Luna against their GPT-5.6 counterparts' promotional pricing. Succeeded by GPT-6.1 Sol on 2026-09-29, and still listed in OpenAI's model catalog with no deprecation notice.","url":"/models/#gpt-6-sol","keywords":"Multimodal closed"},{"type":"Model","title":"OpenAI GPT-6 Luna","description":"OpenAI's most cost-efficient GPT-6 model, aimed at focused, high-volume work at $0.10/$0.50 per million input/output tokens. Takes text and image input with a 1.05M-token context window (922K maximum input, 128K maximum output). It is the GPT-6 model offered to ChatGPT Free and Go users in the desktop app, alongside the paid plans.","url":"/models/#gpt-6-luna","keywords":"Multimodal closed"},{"type":"Model","title":"OpenAI GPT-5.6","description":"OpenAI's GPT-5.6 family, shipping in three tiers - Luna, Terra, and Sol - from most cost-efficient to most capable, with Sol positioned at launch as OpenAI's strongest coding and vision model. The tiers are not deprecated and remain available in the API, but OpenAI's model catalog now leads with the GPT-6 family of Astra, Sol, and Luna.","url":"/models/#gpt-5-6","keywords":"Multimodal closed"},{"type":"Model","title":"OpenAI gpt-oss-120b","description":"OpenAI's larger open-weight model (about 117B parameters), released under Apache 2.0 for local and self-hosted use with a focus on reasoning tasks.","url":"/models/#gpt-oss-120b","keywords":"Apache 2.0 Text open weights"},{"type":"Model","title":"OpenAI gpt-oss-20b","description":"OpenAI's smaller open-weight model (about 22B parameters) under Apache 2.0, designed to run efficiently on modest hardware.","url":"/models/#gpt-oss-20b","keywords":"Apache 2.0 Text open weights"},{"type":"Model","title":"Google DeepMind Gemini 4 Argon","description":"The first model in Google's Gemini 4 generation, announced for complex, long-horizon workflows across real-world software engineering, enterprise knowledge work such as legal and finance, and cyber defense. It raises the output token limit to 1M tokens, up from 64K on earlier Gemini models, which Google says lets it generate hundreds of thousands of tokens in a single trajectory. Announced at an introductory price of $2/$10 per million input/output tokens, rising to $4/$20 afterwards, with cached input at 95% off the input price. At announcement it was rolling out only to a set of trusted cyber defenders through Google's Fairwind Program, with broader access planned starting with paid API customers and Google AI Ultra subscribers; it is not yet listed in the Gemini API model table. Google reports 77.9% on DeepSWE v1.1.","url":"/models/#gemini-4-argon","keywords":"Multimodal closed"},{"type":"Model","title":"Google DeepMind Gemini 3.8 Flash","description":"Google's most intelligent Flash model, built for long-horizon software engineering, autonomous agents, and multi-step reasoning, taking text, image, video, audio, and PDF input with a 1M-token context. Stable in the Gemini API and available in Vertex AI and the Gemini app; a separate Gemini 3.8 Flash Cyber variant for vulnerability detection and patching is limited to trusted defenders through Google's Fairwind Program.","url":"/models/#gemini-3-8-flash","keywords":"Multimodal closed"},{"type":"Model","title":"Google DeepMind Gemini 3.7 Flash","description":"Google's previous Flash workhorse, a high-throughput model tuned for coding and agentic workflows, succeeded by Gemini 3.8 Flash on 2026-09-02. Google says it remains fully supported for efficiency-first workloads.","url":"/models/#gemini-3-7-flash","keywords":"Multimodal closed"},{"type":"Model","title":"Google DeepMind Gemini 3.5 Flash-Lite","description":"The fastest, lowest-cost model in Google's Gemini 3.5 line, aimed at high-volume, latency-sensitive workloads.","url":"/models/#gemini-3-5-flash-lite","keywords":"Multimodal closed"},{"type":"Model","title":"Google DeepMind Gemma 4","description":"Google DeepMind's open-weight family (E2B, E4B, 26B MoE, and 31B dense) built from the same research as Gemini 3, now shipped under Apache 2.0. Handles text, images, audio, and video with up to 256K context.","url":"/models/#gemma-4","keywords":"Apache 2.0 Multimodal open weights"},{"type":"Model","title":"Meta AI Muse Spark","description":"Meta Superintelligence Labs' proprietary flagship, a multimodal reasoning model that powers the Meta AI assistant. Closed-weight and offered through a paid API; the latest version, Muse Spark 1.3, shipped on 2026-09-02 with improved agentic and coding performance and is available through the Muse Code coding agent and the Meta Model API.","url":"/models/#muse-spark","keywords":"Multimodal closed"},{"type":"Model","title":"Meta AI Muse Glimmer 30B","description":"Meta's first open-weight model since Llama 4 and the first open release from Meta Superintelligence Labs. A roughly 29.6B-parameter dense model with text and image input, distilled from the Muse Spark flagship and tuned for on-device agentic use, with a 128K-token context window.","url":"/models/#muse-glimmer-30b","keywords":"Apache 2.0 Multimodal open weights"},{"type":"Model","title":"Meta AI Llama 4 Maverick","description":"Meta's larger open-weight Llama 4 model, a mixture-of-experts design (about 400B total, 17B active parameters) with native text-and-image input.","url":"/models/#llama-4-maverick","keywords":"Llama 4 Community License Multimodal open weights"},{"type":"Model","title":"Meta AI Llama 4 Scout","description":"The smaller Llama 4 model that fits on a single high-end GPU, with 17B active parameters and a very long context window.","url":"/models/#llama-4-scout","keywords":"Llama 4 Community License Multimodal open weights"},{"type":"Model","title":"DeepSeek DeepSeek-V4.1-Flash","description":"The smallest model in DeepSeek's new V4.1 architecture family, a mixture-of-experts model with 552B backbone parameters that activates about 8B per token during prefill and 16B during decode, using a causal encoder-decoder design. MIT-licensed with native image understanding and contexts of up to 1M tokens; it replaced V4-Flash in DeepSeek's API.","url":"/models/#deepseek-v4-1-flash","keywords":"MIT Multimodal open weights"},{"type":"Model","title":"DeepSeek DeepSeek-V4-Pro","description":"DeepSeek's largest V4 open-weight model, a 1.6T-parameter mixture-of-experts release under the MIT license with image input. DeepSeek says V4.1-Flash now outperforms it. DeepSeek first said its API would route V4-Pro requests to V4.1-Flash from 2026-09-14 until V4.1-Pro launches, but later said that, in response to user demand, it will keep serving V4-Pro after that date with billing unchanged; the weights remain available.","url":"/models/#deepseek-v4-pro","keywords":"MIT Multimodal open weights"},{"type":"Model","title":"Alibaba (Qwen) Qwen3.8-Max","description":"Alibaba's flagship Qwen model, a 2.4T-parameter mixture-of-experts system (about 95B active) with a 1M-token context and native text-plus-vision input. The full multimodal Max is API-only; Alibaba released the underlying text-only base checkpoint (Qwen3.8-2.4T-A95B) as open weights on 2026-08-12 under the Qwen3.8-Max license.","url":"/models/#qwen3-8-max","keywords":"Multimodal closed"},{"type":"Model","title":"Alibaba (Qwen) Qwen3.8-27B","description":"An open-weight (Apache 2.0) Qwen model aimed at coding and long-horizon agentic work, with a vision encoder for text, image, and video input and a 256K-token context window (extensible to about 1M). Sized at roughly 27B parameters to run on a single high-end GPU.","url":"/models/#qwen3-8-27b","keywords":"Apache 2.0 Multimodal open weights"},{"type":"Model","title":"Alibaba (Qwen) Qwen3.8-Flash-Next","description":"An open-weight experimental preview of the architecture Qwen says will underpin Qwen4, released as a mixture-of-experts model with about 125B total parameters and roughly 6B activated per token, alongside a 51B n-gram embedding table and a 4B multi-token-prediction module. Pairs Gated DeltaNet with Qwen Sparse Attention in a hybrid attention design across 48 layers and 512 experts. Takes text, image, and video input and returns text, with a 262,144-token native context window extensible to about 1M tokens.","url":"/models/#qwen3-8-flash-next","keywords":"Qwen Community License 1.0 Multimodal open weights"},{"type":"Model","title":"Mistral AI Mistral Medium 3.5","description":"Mistral's first flagship merged model, a dense 128B-parameter model that handles instruction-following, reasoning, and coding in a single set of weights, with text and image input and a 256K-token context window. It became the default model in Mistral's assistant, which was renamed from Le Chat to Vibe in August 2026, and replaced Devstral 2 in the Vibe coding agent. The open weights use a modified MIT license that excludes companies with more than $20 million in monthly revenue, which must get a commercial license from Mistral or use its hosted services.","url":"/models/#mistral-medium-3-5","keywords":"Modified MIT License Multimodal open weights"},{"type":"Model","title":"Mistral AI Mistral Large 3","description":"A 675B-parameter sparse mixture-of-experts open-weight model (41B active) from Mistral under Apache 2.0, with an added vision encoder for image understanding. Mistral's docs still list it as a general-purpose open-weight model alongside its newer flagship, Mistral Medium 3.5.","url":"/models/#mistral-large-3","keywords":"Apache 2.0 Multimodal open weights"},{"type":"Model","title":"SpaceXAI Grok 4.7","description":"SpaceXAI's frontier model for coding, agentic tasks, and knowledge work. SpaceXAI says it uses a new, larger base model than Grok 4.6 and was trained with a longer reinforcement learning run on a harder mix of tasks, weighted toward problems that take many hours to complete. Closed-weight, with a 500K-token context window and text-and-image input. It shipped with what SpaceXAI describes as an entirely new safeguard stack and its strongest results to date on refusals and jailbreak resistance.","url":"/models/#grok-4-7","keywords":"Multimodal closed"},{"type":"Model","title":"SpaceXAI Grok 4.6","description":"SpaceXAI's 1.5T-parameter frontier model, a refinement of Grok 4.5 through improved fine-tuning and reinforcement learning rather than a scale increase. Closed-weight, and succeeded by Grok 4.7 on 2026-09-21 but still listed in SpaceXAI's model and pricing tables alongside it.","url":"/models/#grok-4-6","keywords":"Multimodal closed"},{"type":"Model","title":"MiniMax MiniMax M3","description":"MiniMax's latest M-series model for agentic reasoning, tool use, and coding, with about 428B total parameters of which about 23B are activated. Natively multimodal from the first step of training, it takes text, image, and video input, can operate a desktop computer, and has a 1M-token context window, served by a MiniMax Sparse Attention design that MiniMax says speeds up long-context processing. The MiniMax Community License requires separate written authorization from MiniMax for products with more than $20 million in yearly revenue.","url":"/models/#minimax-m3","keywords":"MiniMax Community License Multimodal open weights"},{"type":"Model","title":"Moonshot AI Kimi K3","description":"Moonshot AI's open-weight mixture-of-experts model with 2.8T total parameters (104B active) and a 1M-token context, among the largest open models released. Multimodal via a native vision encoder.","url":"/models/#kimi-k3","keywords":"Kimi K3 License Multimodal open weights"},{"type":"Model","title":"Z.ai (Zhipu AI) GLM-5.3","description":"Z.ai's flagship open-weight model, a mixture-of-experts release with about 744B total and 40B active parameters that uses the same base model as GLM-5.2, with all gains coming from post-training. Z.ai highlights its coding and emergent cybersecurity capabilities. The GLM-5.3 License is permissive but requires Model-as-a-Service operators with more than $10 billion in annual revenue to pass a Z.ai security review before commercial use.","url":"/models/#glm-5-3","keywords":"GLM-5.3 License Text open weights"},{"type":"Model","title":"Z.ai (Zhipu AI) GLM-5.3-Flash","description":"Z.ai's open-weight (MIT) model with 320B total parameters and about 18B active per token, pairing mixture-of-experts routing with a hybrid sparse-and-linear attention design. The first natively multimodal model in the GLM-5 series, taking text and image input with a 1M-token context window; previewed anonymously as \"Ox Alpha\" and served entirely on domestic Chinese chips.","url":"/models/#glm-5-3-flash","keywords":"MIT Multimodal open weights"},{"type":"Model","title":"NVIDIA NVIDIA Nemotron 3.5 Lightning","description":"NVIDIA's open-weight mixture-of-experts model (about 30B total parameters, roughly 3B active) built for high-volume, long-running agent workloads such as coding, security triage, and customer support. Released with weights, training data, and recipes under the permissive OpenMDW-1.1 license.","url":"/models/#nvidia-nemotron-3-5-lightning","keywords":"OpenMDW-1.1 Text open weights"},{"type":"Lab","title":"Anthropic","description":"SI safety and research company that develops the Claude family of frontier models, with an emphasis on interpretability, alignment, and reliable agentic systems.","url":"/labs/#anthropic"},{"type":"Lab","title":"OpenAI","description":"Research company behind the GPT models and ChatGPT, building general-purpose SI systems along with open-weight releases such as the gpt-oss models.","url":"/labs/#openai"},{"type":"Lab","title":"Google DeepMind","description":"Google's SI research lab, formed in 2023 by merging DeepMind (founded 2010) and Google Brain. Develops the Gemini models and the open-weight Gemma family.","url":"/labs/#google-deepmind"},{"type":"Lab","title":"Meta AI","description":"Meta's SI division, tracing to FAIR (founded 2013) and now organized under Meta Superintelligence Labs. Builds the proprietary Muse Spark flagship and open-weight models, including the Llama family and the newer Muse Glimmer.","url":"/labs/#meta-ai"},{"type":"Lab","title":"SpaceXAI","description":"Elon Musk's SI company, developer of the Grok models. Founded as xAI in 2023, it merged with SpaceX and rebranded to SpaceXAI in July 2026; its models are trained on the Colossus supercomputing clusters and integrated with the X platform.","url":"/labs/#xai"},{"type":"Lab","title":"Mistral AI","description":"Paris-based lab known for efficient open-weight models, including the Mistral and Ministral families, and for Vibe, its unified assistant and coding agent, renamed from Le Chat in August 2026.","url":"/labs/#mistral-ai"},{"type":"Lab","title":"DeepSeek","description":"Hangzhou-based lab backed by the High-Flyer hedge fund, known for MIT-licensed open-weight mixture-of-experts models such as the DeepSeek V4 family.","url":"/labs/#deepseek"},{"type":"Lab","title":"Alibaba (Qwen)","description":"Alibaba's large-model team behind the Qwen (Tongyi Qianwen) family, first released in 2023, spanning frontier Max-class models and open-weight releases.","url":"/labs/#alibaba-qwen"},{"type":"Lab","title":"Moonshot AI","description":"Beijing-based startup that builds the Kimi chatbot and the Kimi series of large models, including large open-weight mixture-of-experts releases.","url":"/labs/#moonshot-ai"},{"type":"Lab","title":"MiniMax","description":"Shanghai-based SI company, publicly listed in Hong Kong in 2026, developing the MiniMax model series for agentic, coding, and multimodal tasks.","url":"/labs/#minimax"},{"type":"Lab","title":"Z.ai (Zhipu AI)","description":"Beijing-based lab that grew out of Tsinghua University, developer of the GLM (General Language Model) family in both API and open-weight form. Formerly known as Zhipu AI, it rebranded to Z.ai and in January 2026 became the first of China's leading 'AI tiger' foundation-model startups to complete a public listing, via a Hong Kong IPO.","url":"/labs/#zhipu-ai"},{"type":"Lab","title":"Cohere","description":"Toronto-based enterprise SI company building the Command model family and retrieval tooling aimed at regulated industries and the public sector.","url":"/labs/#cohere"},{"type":"Lab","title":"Allen Institute for AI (Ai2)","description":"Seattle-based nonprofit founded by Paul Allen, known for fully open models such as the OLMo family, publishing weights, data, and training code.","url":"/labs/#ai2"},{"type":"Lab","title":"NVIDIA","description":"The dominant maker of SI accelerator hardware (GPUs), whose research organization also develops open models, datasets, and tools, including the open-weight Nemotron family for reasoning and agentic tasks.","url":"/labs/#nvidia"},{"type":"Tool","title":"Claude","description":"General-purpose SI assistant from Anthropic for conversation, writing, coding, and document analysis.","url":"/tools/#claude","keywords":"Chat apps"},{"type":"Tool","title":"ChatGPT","description":"OpenAI's conversational assistant with text, image, voice, and web-browsing features.","url":"/tools/#chatgpt","keywords":"Chat apps"},{"type":"Tool","title":"Gemini","description":"Google's SI assistant, integrated with Google Search and Workspace apps.","url":"/tools/#gemini","keywords":"Chat apps"},{"type":"Tool","title":"Microsoft Copilot","description":"Microsoft's SI assistant for chat, web search, and image generation, integrated across Windows and Microsoft 365.","url":"/tools/#microsoft-copilot","keywords":"Chat apps"},{"type":"Tool","title":"Perplexity","description":"SI answer engine that responds to questions with cited web sources.","url":"/tools/#perplexity","keywords":"Chat apps"},{"type":"Tool","title":"Grok","description":"Conversational SI from SpaceXAI with real-time access to web and X content.","url":"/tools/#grok","keywords":"Chat apps"},{"type":"Tool","title":"DeepSeek","description":"Chat assistant from DeepSeek AI that runs the company's open-weight models.","url":"/tools/#deepseek","keywords":"Chat apps"},{"type":"Tool","title":"Mistral Vibe","description":"Mistral's assistant and coding agent in one product, covering conversational chat, longer-horizon work tasks, and coding; renamed from Le Chat in August 2026.","url":"/tools/#mistral-vibe","keywords":"Chat apps"},{"type":"Tool","title":"LM Studio","description":"Desktop application for discovering, downloading, and running local LLMs with a chat UI and an OpenAI-compatible local server.","url":"/tools/#lm-studio","keywords":"Local runners"},{"type":"Tool","title":"Ollama","description":"Tool for downloading and running open-weight models locally, with a command line and a local API; running models on your own hardware is free, with optional paid plans.","url":"/tools/#ollama","keywords":"Local runners"},{"type":"Tool","title":"llama.cpp","description":"C/C++ library and command-line tools for running LLM inference locally across a wide range of hardware.","url":"/tools/#llama-cpp","keywords":"Local runners"},{"type":"Tool","title":"Jan","description":"Open-source desktop app to run LLMs offline, with local chat history and an OpenAI-compatible server.","url":"/tools/#jan","keywords":"Local runners"},{"type":"Tool","title":"GPT4All","description":"Desktop app from Nomic AI for running open-source LLMs privately on your own device, including local document chat.","url":"/tools/#gpt4all","keywords":"Local runners"},{"type":"Tool","title":"Claude Code","description":"Anthropic's agentic coding tool that works in the terminal, IDEs, and the web to build, debug, and ship code across a codebase.","url":"/tools/#claude-code","keywords":"Coding agents & dev tools"},{"type":"Tool","title":"Cursor","description":"SI-native code editor, built as a VS Code fork, with in-editor chat, autocomplete, and agentic multi-file editing.","url":"/tools/#cursor","keywords":"Coding agents & dev tools"},{"type":"Tool","title":"GitHub Copilot","description":"SI pair-programmer from GitHub offering code completion and chat across major IDEs and the command line.","url":"/tools/#github-copilot","keywords":"Coding agents & dev tools"},{"type":"Tool","title":"OpenAI Codex CLI","description":"OpenAI's open-source terminal coding agent, running tasks in a local sandbox on macOS, Windows, and Linux. It is one surface of a wider Codex family that also includes an IDE extension, a desktop app, and Codex Cloud for running tasks in cloud environments; the CLI, IDE extension, and web surfaces begin at the Plus plan, while Free and Go get Codex in the desktop app only.","url":"/tools/#openai-codex-cli","keywords":"Coding agents & dev tools"},{"type":"Tool","title":"Google Antigravity","description":"Google's agent-first development platform, built around Gemini models: a standalone desktop application, extensions for other editors, a CLI, and an SDK.","url":"/tools/#google-antigravity","keywords":"Coding agents & dev tools"},{"type":"Tool","title":"Aider","description":"Open-source command-line coding assistant that pairs with your local git repository; you bring your own model API key.","url":"/tools/#aider","keywords":"Coding agents & dev tools"},{"type":"Tool","title":"Cline","description":"Open-source autonomous coding agent that runs as a sidebar in VS Code and other editors or from the command line; you bring your own model API key.","url":"/tools/#cline","keywords":"Coding agents & dev tools"},{"type":"Tool","title":"Zed","description":"Open-source, high-performance code editor written in Rust with built-in SI assistance and collaboration features.","url":"/tools/#zed","keywords":"Coding agents & dev tools"},{"type":"Tool","title":"Midjourney","description":"Subscription image-generation service, accessible through a web app and Discord.","url":"/tools/#midjourney","keywords":"Creative tools"},{"type":"Tool","title":"Google Veo","description":"Google DeepMind's text-to-video and image-to-video model with native audio generation, available through Gemini, Flow, and the Gemini API.","url":"/tools/#google-veo","keywords":"Creative tools"},{"type":"Tool","title":"Google Imagen","description":"Google DeepMind's text-to-image model, available through Gemini, Google AI Studio, and the Gemini API.","url":"/tools/#google-imagen","keywords":"Creative tools"},{"type":"Tool","title":"ElevenLabs","description":"SI audio platform for text-to-speech, voice cloning, dubbing, and music generation.","url":"/tools/#elevenlabs","keywords":"Creative tools"},{"type":"Tool","title":"Suno","description":"SI music generator that creates songs with vocals and instrumentation from text prompts.","url":"/tools/#suno","keywords":"Creative tools"},{"type":"Tool","title":"Runway","description":"SI platform for video generation and editing, including text-to-video and image-to-video tools.","url":"/tools/#runway","keywords":"Creative tools"},{"type":"Tool","title":"Ideogram","description":"SI image generator noted for rendering legible text within images.","url":"/tools/#ideogram","keywords":"Creative tools"},{"type":"Tool","title":"Claude Agent SDK","description":"Anthropic's SDK for building agents on the same tools, agent loop, and context management that power Claude Code; formerly the Claude Code SDK.","url":"/tools/#claude-agent-sdk","keywords":"Agent frameworks & libraries"},{"type":"Tool","title":"LangChain","description":"Open-source framework for building applications and agents with LLMs, with integrations across many model and tool providers.","url":"/tools/#langchain","keywords":"Agent frameworks & libraries"},{"type":"Tool","title":"LangGraph","description":"Open-source library from the LangChain team for building stateful, multi-step agent workflows modeled as graphs.","url":"/tools/#langgraph","keywords":"Agent frameworks & libraries"},{"type":"Tool","title":"CrewAI","description":"Open-source Python framework for orchestrating role-based, collaborating SI agents.","url":"/tools/#crewai","keywords":"Agent frameworks & libraries"},{"type":"Tool","title":"OpenAI Agents SDK","description":"OpenAI's lightweight open-source SDK for building multi-agent workflows with handoffs, guardrails, and tracing.","url":"/tools/#openai-agents-sdk","keywords":"Agent frameworks & libraries"},{"type":"Tool","title":"LlamaIndex","description":"Open-source data framework for connecting LLMs and agents to external data sources for retrieval and workflows.","url":"/tools/#llamaindex","keywords":"Agent frameworks & libraries"},{"type":"Tool","title":"Google Agent Development Kit (ADK)","description":"Google's open-source Agent Development Kit for building, testing, and deploying agents, with support for multiple languages.","url":"/tools/#google-adk","keywords":"Agent frameworks & libraries"},{"type":"Glossary","title":"Agent","description":"A software system that uses an SI model to take actions on its own toward a goal, such as browsing the web, running code, or using tools, rather than just answering a single question. Agents typically work in a loop: decide, act, observe the result, and try again.","url":"/glossary/#agent"},{"type":"Glossary","title":"AGI (Artificial General Intelligence)","description":"General SI, long called artificial general intelligence (AGI): a hypothetical system that could understand and perform any intellectual task a human can, across many domains rather than one narrow specialty. There is no agreed-upon test for when it would be reached, and experts disagree on whether or when it will happen.","url":"/glossary/#agi-artificial-general-intelligence"},{"type":"Glossary","title":"AI vs SI","description":"Two names for the same technology. What changed is vocabulary: on September 29, 2026 a US executive order directed federal departments and agencies to say \"Super Intelligence\" (SI) instead of \"artificial intelligence\" (AI) in official communications. What didn't change is the technology or the law: models work exactly as before, and statutes, state laws, contracts, the EU AI Act and most vendor documentation still say \"AI\", so many organizations now use both terms.","url":"/glossary/#ai-vs-si"},{"type":"Glossary","title":"Alignment","description":"The effort to make SI systems pursue the goals and values their designers and users actually intend, and to behave safely and honestly. Misalignment is when a system optimizes for something subtly different from what people want.","url":"/glossary/#alignment"},{"type":"Glossary","title":"Artificial intelligence (AI)","description":"The name the field carried for seven decades, coined at the 1956 Dartmouth workshop that is widely regarded as its founding. In 2026 the US government moved to a new term: a September 29 executive order directed federal departments and agencies to use \"Super Intelligence\" (SI) in place of \"artificial intelligence\" in official communications, on the stated grounds that \"artificial\" makes the technology sound fake. AI and SI name the same technology, and the older term remains in laws, contracts, the EU AI Act and most vendor documentation.","url":"/glossary/#artificial-intelligence-ai"},{"type":"Glossary","title":"ASI (Artificial Superintelligence)","description":"A hypothetical system that would be far more capable than the best humans at essentially everything. It is a step beyond general SI (AGI) and remains speculative. Despite the similar name, it is not what everyday super intelligence (SI) means: SI is the 2026 US federal term for the systems in use today, and none of them is artificial superintelligence.","url":"/glossary/#asi-artificial-superintelligence"},{"type":"Glossary","title":"Attention","description":"A mechanism inside modern SI models that lets them weigh which earlier words or pieces of input matter most when producing the next piece of output. It is the core idea behind the transformer, the architecture used by most large language models.","url":"/glossary/#attention"},{"type":"Glossary","title":"Backpropagation","description":"The core training algorithm for neural networks. It works out how much each internal setting contributed to the model's errors and nudges all of them in the direction that reduces those errors, repeated over huge amounts of data.","url":"/glossary/#backpropagation"},{"type":"Glossary","title":"Benchmark","description":"A standardized test used to measure and compare how well SI models perform on a specific kind of task, such as math, coding, or general knowledge. Benchmarks help track progress, but a high score does not always translate to real-world usefulness.","url":"/glossary/#benchmark"},{"type":"Glossary","title":"Chain of thought","description":"A technique where a model works through a problem step by step in writing before giving its final answer, much like showing your work in math. It often improves accuracy on reasoning-heavy tasks.","url":"/glossary/#chain-of-thought"},{"type":"Glossary","title":"Chatbot","description":"A program you interact with by typing or speaking in everyday language, and that responds conversationally. Modern chatbots like ChatGPT and Claude are powered by large language models.","url":"/glossary/#chatbot"},{"type":"Glossary","title":"Compute","description":"The raw computing power, usually measured as processor time on chips like GPUs, needed to train or run an SI model. More compute generally allows bigger models and more training, and it is one of the main costs and bottlenecks in SI.","url":"/glossary/#compute"},{"type":"Glossary","title":"Context engineering","description":"The practice of curating what goes into a model's context window at each step - instructions, retrieved documents, tool outputs, conversation history, and state - so the model has what it needs and little else. It differs from prompt engineering, which focuses on the wording of the instruction; context engineering treats the whole information environment as a finite budget to be managed, which matters most for long-running agents.","url":"/glossary/#context-engineering"},{"type":"Glossary","title":"Context window","description":"The amount of text, measured in tokens, that a model can take in and consider at once, including both your input and its own response. Anything beyond that limit is dropped, so a larger context window lets a model work with longer documents or conversations.","url":"/glossary/#context-window"},{"type":"Glossary","title":"Deep learning","description":"A type of machine learning that uses neural networks with many layers to learn patterns directly from large amounts of data. It powers most modern SI, including image recognition and large language models.","url":"/glossary/#deep-learning"},{"type":"Glossary","title":"Diffusion model","description":"A type of SI model that generates images or other data by starting with random noise and gradually refining it into a coherent result. It is the technology behind many popular image generators.","url":"/glossary/#diffusion-model"},{"type":"Glossary","title":"Distillation","description":"A technique for training a smaller, cheaper student model to imitate a larger, more capable teacher model. The goal is to keep much of the quality while cutting the cost and speed of running it.","url":"/glossary/#distillation"},{"type":"Glossary","title":"Effort (reasoning effort)","description":"A request setting that tells a model how much work to put into an answer, trading thoroughness against token cost and latency on the same model. It has become a standard control across providers: Anthropic exposes an effort parameter with levels from low to max, and OpenAI exposes a reasoning effort parameter with a comparable range. Because reported benchmark scores depend on the level used, the effort setting is now part of reading a model's results.","url":"/glossary/#effort-reasoning-effort"},{"type":"Glossary","title":"Embedding","description":"A way of representing words, images, or other data as a list of numbers so that items with similar meaning end up close together. Embeddings let software measure similarity and are a building block of search and recommendation systems.","url":"/glossary/#embedding"},{"type":"Glossary","title":"Emergent abilities","description":"Skills that appear in large models but were not present in smaller ones, seeming to switch on once a model reaches a certain size or amount of training. Researchers debate how real and how predictable these jumps actually are.","url":"/glossary/#emergent-abilities"},{"type":"Glossary","title":"Few-shot learning","description":"Giving a model a handful of examples of a task inside the prompt so it can follow the pattern, without any additional training. It contrasts with zero-shot, where no examples are provided.","url":"/glossary/#few-shot-learning"},{"type":"Glossary","title":"Fine-tuning","description":"Taking an already-trained model and training it further on a narrower set of examples so it does better at a specific task or adopts a particular style. It is usually far cheaper than training a model from scratch.","url":"/glossary/#fine-tuning"},{"type":"Glossary","title":"Foundation model","description":"A large model trained on broad data that can be adapted to many different tasks, serving as a base that others build on. Large language models are the best-known example.","url":"/glossary/#foundation-model"},{"type":"Glossary","title":"Frontier model","description":"One of the most capable SI models available at a given time, typically from a leading lab and trained at very large scale. The term comes up often in discussions of SI safety and regulation.","url":"/glossary/#frontier-model"},{"type":"Glossary","title":"Generative AI","description":"Generative SI, still widely called generative AI: systems that create new content, such as text, images, audio, video, or code, rather than just classifying or scoring existing data. Chatbots and image generators are common examples.","url":"/glossary/#generative-ai"},{"type":"Glossary","title":"GPU","description":"Short for graphics processing unit, a type of chip originally built for rendering graphics that turns out to be very good at the parallel math SI models need. GPUs are the main hardware used to train and run modern SI systems.","url":"/glossary/#gpu"},{"type":"Glossary","title":"Guardrails","description":"Rules, filters, or added safety layers that keep an SI system from producing harmful or off-limits outputs. They sit around the model rather than being part of how it fundamentally works.","url":"/glossary/#guardrails"},{"type":"Glossary","title":"Hallucination","description":"When an SI model states something false or made-up as if it were true, often fluently and confidently. It happens because models predict plausible-sounding text rather than looking up verified facts.","url":"/glossary/#hallucination"},{"type":"Glossary","title":"Inference","description":"The act of running a trained model to get an answer, as opposed to training it. Every time you send a prompt and get a response, that is inference.","url":"/glossary/#inference"},{"type":"Glossary","title":"Jailbreak","description":"A prompt or trick designed to get an SI model to bypass its safety rules and produce content it is meant to refuse. Labs continually patch known jailbreaks as they are discovered.","url":"/glossary/#jailbreak"},{"type":"Glossary","title":"Knowledge cutoff","description":"The date after which a model has no built-in knowledge, because its training data only goes up to that point. Events after the cutoff are unknown to the model unless it is given that information or can look it up.","url":"/glossary/#knowledge-cutoff"},{"type":"Glossary","title":"Large language model (LLM)","description":"An SI model trained on huge amounts of text to predict and generate language, which lets it answer questions, write, summarize, and more. Often abbreviated LLM, these models power today's leading chatbots.","url":"/glossary/#large-language-model-llm"},{"type":"Glossary","title":"Latency","description":"The delay between sending a request to an SI model and getting a response back. Lower latency means the system feels faster and more responsive.","url":"/glossary/#latency"},{"type":"Glossary","title":"Machine learning","description":"A branch of SI where systems learn patterns from data and improve with experience, instead of being explicitly programmed with rules for every case. Deep learning is one powerful kind of machine learning.","url":"/glossary/#machine-learning"},{"type":"Glossary","title":"MCP (Model Context Protocol)","description":"An open standard, introduced by Anthropic in late 2024, for connecting SI models to outside tools and data sources in a consistent way. It lets developers plug an assistant into things like files, databases, and apps without custom wiring for each one. Anthropic donated it to the Linux Foundation's Agentic AI Foundation in December 2025, and it is now widely supported across competing SI products.","url":"/glossary/#mcp-model-context-protocol"},{"type":"Glossary","title":"Mixture of experts","description":"A model design that splits the network into many specialized sub-models, or experts, and for each input uses only a few of them. This keeps a model large in total knowledge while making each response cheaper to compute.","url":"/glossary/#mixture-of-experts"},{"type":"Glossary","title":"Multimodal","description":"Describes an SI model that can handle more than one type of input or output, such as text, images, audio, and video, rather than text alone. Most leading models today are multimodal.","url":"/glossary/#multimodal"},{"type":"Glossary","title":"Neural network","description":"A computing system loosely inspired by the brain, made of layers of simple connected units whose numeric settings are tuned during training. It is the basic structure underneath deep learning and modern SI.","url":"/glossary/#neural-network"},{"type":"Glossary","title":"Open weights","description":"When a model's trained parameters are released publicly so anyone can download, run, and adapt it. This differs from fully open source, since the training data and code may not be shared, and from closed models offered only through an online service.","url":"/glossary/#open-weights"},{"type":"Glossary","title":"Overfitting","description":"When a model learns its training examples too closely, including their quirks and noise, and as a result performs worse on new, unseen data. Avoiding it is a central concern in machine learning.","url":"/glossary/#overfitting"},{"type":"Glossary","title":"Parameter","description":"One of the internal numeric values a model adjusts during training; together they store what the model has learned. Modern large models have billions or more, and the count is a rough, imperfect indicator of capacity.","url":"/glossary/#parameter"},{"type":"Glossary","title":"Pre-training","description":"The first and largest training stage, where a model learns general patterns from a huge, broad dataset. For language models this usually means learning to predict the next token across vast amounts of text, with later stages like fine-tuning specializing it.","url":"/glossary/#pre-training"},{"type":"Glossary","title":"Prompt","description":"The input or instruction you give an SI model to get a response, such as a question, a request, or any text you type in. The wording of a prompt can strongly affect the quality of the answer.","url":"/glossary/#prompt"},{"type":"Glossary","title":"Prompt engineering","description":"The practice of carefully wording and structuring prompts to get better, more reliable results from an SI model. It can include giving examples, setting a role, or breaking a task into steps.","url":"/glossary/#prompt-engineering"},{"type":"Glossary","title":"Prompt injection","description":"An attack where hidden or malicious instructions, often buried in a web page or document the model reads, trick it into ignoring its real instructions. It is a major security concern for SI agents that browse or process outside content.","url":"/glossary/#prompt-injection"},{"type":"Glossary","title":"Quantization","description":"A technique that shrinks a model by storing its numbers with less precision, so it uses less memory and runs faster. It usually costs a little accuracy in exchange for being cheaper to run.","url":"/glossary/#quantization"},{"type":"Glossary","title":"RAG (Retrieval-Augmented Generation)","description":"A method where the system first looks up relevant information from an outside source, such as a document collection, and feeds it to the model so answers are grounded in that material. It helps reduce hallucination and lets a model use up-to-date or private data.","url":"/glossary/#rag-retrieval-augmented-generation"},{"type":"Glossary","title":"Reasoning model","description":"A model trained to spend extra effort thinking through a problem step by step before answering, which improves performance on hard math, coding, and logic tasks. OpenAI's o1 and DeepSeek-R1 are well-known examples.","url":"/glossary/#reasoning-model"},{"type":"Glossary","title":"Red teaming","description":"Deliberately probing an SI system to find its weaknesses, harmful outputs, and ways it can be misused, so they can be fixed before release. The idea and name are borrowed from security testing.","url":"/glossary/#red-teaming"},{"type":"Glossary","title":"Reinforcement learning","description":"A training approach where a system learns by trial and error, receiving rewards for good outcomes and penalties for bad ones, and adjusting to earn more reward. It is used both in game-playing SI and in fine-tuning language models.","url":"/glossary/#reinforcement-learning"},{"type":"Glossary","title":"Reward hacking","description":"When a system trained against a reward or score finds a way to raise that number without doing the thing the number was meant to measure, such as writing code that satisfies a test rather than fixing the underlying bug. Also called specification gaming, it is a central difficulty in training and evaluating models, because most practical objectives are imperfect proxies for what designers actually want.","url":"/glossary/#reward-hacking"},{"type":"Glossary","title":"RLHF (Reinforcement Learning from Human Feedback)","description":"A training method where people rate or rank a model's responses, and those judgments are used to steer the model toward more helpful and appropriate answers. It was central to making chatbots like ChatGPT feel useful and polite.","url":"/glossary/#rlhf-reinforcement-learning-from-human-feedback"},{"type":"Glossary","title":"Scaling laws","description":"Observed, fairly predictable relationships showing that model performance tends to improve as you increase model size, training data, and compute. They have guided the industry's push toward ever-larger models.","url":"/glossary/#scaling-laws"},{"type":"Glossary","title":"Speculative decoding","description":"A speed-up technique where a small, fast model drafts several tokens ahead and a larger model checks them in one pass, accepting the ones it agrees with. It produces the same output as the big model alone but faster.","url":"/glossary/#speculative-decoding"},{"type":"Glossary","title":"Super Intelligence (SI)","description":"The term the US government adopted in 2026 for what was called artificial intelligence (AI): software systems such as large language models, chatbots and agents that learn from data to write, code, analyze, and generate images and other media. In computer science, this everyday SI is distinct from artificial superintelligence (ASI), an older and still hypothetical idea of a system that outperforms humans at most intellectual tasks; today's SI systems are not that.","url":"/glossary/#super-intelligence-si"},{"type":"Glossary","title":"Super Intelligence executive order","description":"A White House executive order issued on September 29, 2026 that directs federal departments and agencies to use the term \"Super Intelligence\" (SI) instead of \"artificial intelligence\" in official communications. Its stated reasoning is that \"artificial\" makes the technology sound fake, while \"Super Intelligence\" better captures its promise and capabilities. The order changes vocabulary only, not the technology or existing law.","url":"/glossary/#super-intelligence-executive-order"},{"type":"Glossary","title":"Synthetic data","description":"Training data generated by a computer or another SI model, rather than collected from the real world. It can fill gaps where real data is scarce, expensive, or sensitive, though it carries risks if the generated data is flawed.","url":"/glossary/#synthetic-data"},{"type":"Glossary","title":"System prompt","description":"A behind-the-scenes instruction that sets an SI assistant's role, tone, and rules before the conversation with the user begins. Users usually do not see it, but it shapes how the assistant responds.","url":"/glossary/#system-prompt"},{"type":"Glossary","title":"Temperature","description":"A setting that controls how random or predictable a model's output is. Lower values make responses more focused and repeatable; higher values make them more varied and creative.","url":"/glossary/#temperature"},{"type":"Glossary","title":"Test-time compute","description":"The idea of letting a model use more computation while answering, for example by reasoning longer or trying multiple approaches, to get better results. Reasoning models rely heavily on this.","url":"/glossary/#test-time-compute"},{"type":"Glossary","title":"Token","description":"A chunk of text, often a word or part of a word, that a language model reads and generates one at a time. Models measure input length, output length, and pricing in tokens.","url":"/glossary/#token"},{"type":"Glossary","title":"Tool use","description":"When an SI model calls external tools, such as a calculator, a search engine, or a piece of software, to do things it cannot do reliably on its own. It lets models fetch fresh information and take real actions.","url":"/glossary/#tool-use"},{"type":"Glossary","title":"Training data","description":"The collection of examples a model learns from during training. Its size, quality, and biases strongly shape what the model knows and how it behaves.","url":"/glossary/#training-data"},{"type":"Glossary","title":"Training run","description":"A single, complete session of training a model from start to finish on a chosen dataset and setup. For large models, one run can take weeks and cost millions of dollars.","url":"/glossary/#training-run"},{"type":"Glossary","title":"Transformer","description":"The neural network architecture, introduced in 2017, that underlies most modern language models. Its key innovation, the attention mechanism, lets it handle long stretches of text efficiently and learn relationships between distant words.","url":"/glossary/#transformer"},{"type":"Glossary","title":"Turing test","description":"A thought experiment proposed by Alan Turing in 1950, in which a machine passes if a human judge chatting with it cannot reliably tell it apart from a person. It shaped decades of debate about machine intelligence, though many now see it as a limited measure.","url":"/glossary/#turing-test"},{"type":"Glossary","title":"Vector database","description":"A specialized database that stores embeddings and can quickly find the items most similar in meaning to a query. It is a common building block for search and for retrieval-augmented generation.","url":"/glossary/#vector-database"},{"type":"Glossary","title":"Vibe coding","description":"An informal style of programming, popularized in 2025, where a person describes what they want in plain language and lets an SI model generate the code, guiding it by feel and results rather than writing much code by hand. The term was coined by researcher Andrej Karpathy.","url":"/glossary/#vibe-coding"},{"type":"Glossary","title":"World model","description":"An SI system's internal representation of how the world works, which it can use to predict what happens next or plan ahead. How much today's models truly have one is an open research question.","url":"/glossary/#world-model"},{"type":"Glossary","title":"Zero-shot learning","description":"When a model performs a task it was given no examples for, relying only on its general training and the instructions in the prompt. It contrasts with few-shot, where a few examples are provided.","url":"/glossary/#zero-shot-learning"},{"type":"Coverage","title":"SI Models news","description":"New super intelligence (SI) model releases, upgrades, and capability jumps — frontier labs and open weights alike.","url":"/topic/models-news/"},{"type":"Coverage","title":"SI Agents & Tools news","description":"SI agents for coding and computer use, agent frameworks, and the SI tools people actually build with.","url":"/topic/agents/"},{"type":"Coverage","title":"SI Research news","description":"Papers, benchmarks, training techniques, and measurable progress toward general SI (AGI).","url":"/topic/research/"},{"type":"Coverage","title":"SI Industry news","description":"The SI labs, the money, the partnerships, and the competition to build general SI.","url":"/topic/industry/"},{"type":"Coverage","title":"SI Safety & Policy news","description":"SI safety and alignment, evaluations, governance, and how governments are responding to frontier SI.","url":"/topic/safety/"},{"type":"Coverage","title":"Open-Source SI news","description":"Open-weight SI models, local inference, and the tooling that runs SI on your own hardware.","url":"/topic/open-source/"},{"type":"Coverage","title":"SI Hardware & Compute news","description":"GPUs, accelerators, datacenters, and the compute buildout behind the SI race.","url":"/topic/hardware/"},{"type":"Coverage","title":"SI Business news","description":"Funding rounds, valuations, enterprise SI adoption, and SI product economics.","url":"/topic/business/"}]