Desk

Your next step

Start here

Take the 15-question test

We will not invent a level. About four minutes. Then we tell you the first official module to open.

Find my level

Send a friend the free test

Wire · scanned 2026-10-17

The landscape moves every week. Sometimes every day. The Desk shows the items that change a path — not a firehose.

confirmed · 2026-10-13 · L1–L4 workflow

Gemini 4 Argon is generally available — Gemini app for Google AI Pro/Ultra; API $2/$10 MTok intro — not a free course

Tuesday: Google made Gemini 4 Argon, its frontier model for coding, agentic work, math, and science, generally available (official Oct 13, 2026). Consumers: Gemini app and AI Mode in Search for Google AI Pro and Ultra subscribers (not the free tier). Developers: Gemini API paid tier in AI Studio and Android Studio, model ID gemini-4-argon, 1M unified input-output token limit; Ultra subscribers can also try it in Google Antigravity and AI Studio. Business: Gemini for business, the API, and Antigravity. Intro API price $2 / $10 per MTok (cached input 95% off) through Nov 30, 2026, then $4 / $20. Gemini app limits stay compute-based (refresh every 5 hours up to a weekly cap; Ultra 5x or 20x Pro). Fairwind (no cyber guardrails) and the 10,000-scientist free Antigravity tier are application programs, not PATH doors. This is a paid model surface, not a free course — Grow with Google stays the door: https://grow.google/ai/. Do not catalog. Official: https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/. API changelog: https://ai.google.dev/gemini-api/docs/changelog. Limits: https://support.google.com/gemini/answer/16275805. App: https://gemini.google/release-notes/.

confirmed · 2026-10-07 · L1–L4 workflow

GPT-6 reaches ChatGPT Chat for everyone — Sol on paid tiers, Luna on Free/Go, plus Intelligent UI

Wednesday: OpenAI brought GPT-6 to the Chat tab in ChatGPT (official Oct 7, 2026). Plus / Pro / Business / Enterprise get GPT-6 Sol starting Oct 7; Free and Go get GPT-6 Luna starting Oct 8 (rolling out globally; Enterprise depends on admin settings). New Intelligent UI: answers can include visuals, buttons, forms, charts, and small interactive tools; ChatGPT can start answering while it keeps thinking (GPT-6 Instant web-search answers start ~44% sooner than GPT-5.6 Instant, per OpenAI). Work and Codex models are not changing in this release. Help Center: GPT-6 replaces Latest in the model menu; Plus gets Medium/High, Pro/Business/Enterprise add Extra High and Pro (GPT-6 Pro on Astra); Intelligent UI has no separate quota; Free/Go keep unlimited everyday text chats (tools have separate limits). Re-checked Oct 17: the Help Center (last updated about Oct 9) still describes Free/Go as GPT-5.6 Luna while OpenAI’s announcement and Oct 7 release notes say GPT-6 Luna from Oct 8 — follow OpenAI’s own pages, do not invent plan rules. This matters for PATH because free learners now meet a new default in Chat; it is not a course — OpenAI Academy stays the free door: https://academy.openai.com/. Official: https://openai.com/index/gpt-6-for-everyone/. Help: https://help.openai.com/en/articles/20001354-gpt-5-6-in-chatgpt. Release notes: https://help.openai.com/en/articles/6825453-chatgpt-release-notes. News: https://openai.com/news/.

confirmed · 2026-10-07 · L1–L4 workflow

Claude Haiku 5.5 is generally available — fastest, cheapest Claude; $0.10/$0.50 MTok; Sonnet 5.5 cache reads cut 50% — not a free course

Wednesday: Anthropic shipped Claude Haiku 5.5, the small model in the Claude 5.5 family (official Oct 7, 2026). Anthropic calls it its fastest model to date at standard speed and says it costs about 75% less to run than Haiku 4.5 on average. List $0.10 / $0.50 per MTok for prompts up to 100k tokens ($0.50 / $2.50 over 100k); cache reads $0.01. Model ID claude-haiku-5-5 on Claude Platform; also Amazon Web Services, Google Cloud, and Microsoft Azure; 1M-token context. Same day: Sonnet 5.5 cache reads drop 50% to $0.10 per MTok (Anthropic says ~20% cheaper on most agentic work), and Max and Team subscribers get a new monthly Claude Platform API credit (Max 5x $100, Max 20x $200, Team $20/$100 per Standard/Premium seat pooled up to $500; Free, Pro, and Enterprise not eligible; does not cover interactive Claude Code or extra usage). Computer use and browser use SDK support in beta. This is a paid model / API surface, not a free course — Anthropic Academy stays the free door: https://anthropic.skilljar.com/. Do not invent plan-inclusion rules beyond Anthropic’s own pages. Official: https://www.anthropic.com/claude-haiku-5-5. Docs: https://platform.claude.com/docs/en/models/haiku-5-5/overview. API credits: https://support.claude.com/en/articles/17154008. Release notes: https://support.claude.com/en/articles/12138966-release-notes.

confirmed · 2026-09-29 · L1–L4 workflow

GPT-6.1 Sol is available — near-Astra Work/Codex upgrade; API $2/$10 MTok, cached input $0.10 — not a free course

Tuesday: OpenAI shipped GPT-6.1 Sol, an upgrade to GPT-6 Sol (official Product Sep 29, 2026 / DevDay window). Nearly matches GPT-6 Astra on agentic coding, computer use, and professional work at about one-fifth of Astra’s standard input/output token prices. Cached input $0.10 per MTok (95% less than standard input; 50% less than GPT-6 Sol cached input). Available starting today to Plus / Pro / Business / Enterprise / Edu in ChatGPT Work and Codex; not yet in Chat. API model ID gpt-6.1-sol; list $2 / $10 per MTok (cached input $0.10). GPT-6.1 Sol Ultrafast said to follow in coming days (up to 8× faster token generation in Codex) — not GA yet. Help Center now notes 6.1 Sol rolling out in Work/Codex; Chat narrative may still describe GPT-5.6 / Astra allowances — Astra Chat caps unchanged as a separate surface. This is a paid model / API surface, not a free course — OpenAI Academy stays the free door: https://academy.openai.com/. Do not invent plan-inclusion rules beyond OpenAI’s own pages. Official: https://openai.com/index/introducing-gpt-6-1-sol/. Safety addendum: https://deploymentsafety.openai.com/gpt-6-1-sol. News: https://openai.com/news/.

confirmed · 2026-09-28 · L1–L4 workflow

Claude Sonnet 5.5 is generally available — faster/cheaper complement to Opus 5.5; $2/$10 MTok — not a free course

Monday: Anthropic shipped Claude Sonnet 5.5, second model in the Claude 5.5 family (official Sep 28, 2026). Clear upgrade over Sonnet 5 — runs 30%+ faster and costs up to 30% less for most work while list price stays $2 / $10 per MTok (cache writes $2.50; cache reads cut to $0.10 on Oct 7, from $0.20). Strongest at well-scoped everyday tasks, bug fixes, and polished documents/slides/spreadsheets; complement to Opus 5.5 for complex judgment work. Model ID claude-sonnet-5-5 on Claude Platform; also Amazon Web Services, Google Cloud, and Microsoft Azure; zero data retention like Opus 5.5 / Sonnet 5. Claude Code and apps default effort Medium (Platform defaults High). Haiku 5.5 now GA (see Oct 7 item). Cyber safeguards comparable to Opus 5 class on this Sonnet; biology safeguards same as Sonnet 5 — Mythos-style trusted programs are not PATH doors. This is a paid model surface, not a free course — Anthropic Academy stays the free door: https://anthropic.skilljar.com/. Do not invent plan-inclusion rules beyond Anthropic’s own pages. Official: https://www.anthropic.com/claude-sonnet-5-5. Docs: https://platform.claude.com/docs/en/models/sonnet-5-5/overview. Newsroom: https://www.anthropic.com/news (Sep 28 Announcements).

confirmed · 2026-09-22 · L1–L4 workflow

GPT-6 Sol and Luna are available — Work/Codex (+ Luna desktop for Free/Go); API 50% cheaper vs 5.6 promo — not a free course

Tuesday: OpenAI expanded the GPT-6 family with GPT-6 Sol and GPT-6 Luna (official Sep 22, 2026). Trained with similar methods as GPT-6 Astra for faster/cheaper tiers. API model IDs gpt-6-sol and gpt-6-luna; list prices Sol $2/$10 and Luna $0.10/$0.50 per MTok — 50% cheaper vs GPT-5.6 Sol/Luna promotional pricing. Available in ChatGPT Work and Codex for Plus / Pro / Business / Enterprise / Edu; Free and Go get Luna in the desktop app; GPT-6 reached Chat on Oct 7 (see Oct 7 item). Companion Product post improves GPT-6 prompt caching (up to 90% cached-input discount). This is a model / API surface, not a free course — OpenAI Academy stays the free door: https://academy.openai.com/. Do not catalog. Prefer the Product post for Sol/Luna GA; GPT-6.1 Sol upgrade now GA for Work/Codex (see Sep 29 item). Help Center Chat narrative may still describe GPT-5.6 Sol/Luna defaults — Astra allowances unchanged per Help. Official: https://openai.com/index/introducing-gpt-6-sol-and-luna/. Caching: https://openai.com/index/better-prompt-caching-for-gpt-6/. Astra Help (unchanged allowances): https://help.openai.com/en/articles/20001354-gpt-5-6-in-chatgpt.

confirmed · 2026-09-22 · L2–L4 workflow

Claude Opus 5.5 is generally available — frontier agentic coding / knowledge work; $4/$20 MTok — not a free course

Tuesday: Anthropic shipped Claude Opus 5.5, first model in the Claude 5.5 family (official Sep 22, 2026). Built for long-running agentic coding and knowledge work; Anthropic says it performs at Claude Fable 5.1 level on most work and costs ~40% less to run than Opus 5 on typical workloads. List price $4 / $20 per MTok (cache reads $0.20). Model ID claude-opus-5-5 on Claude API; also Amazon Bedrock, Google Cloud, Microsoft Foundry / Azure. Adaptive thinking always on; Sonnet 5.5 GA Sep 28 and Haiku 5.5 GA Oct 7 (see items above). Biology / cybersecurity safeguards comparable to Fable 5.1 / Mythos class — Mythos-style trusted programs are not PATH doors. This is a paid model surface, not a free course — Anthropic Academy stays the free door: https://anthropic.skilljar.com/. Do not invent plan-inclusion rules beyond Anthropic’s own pages. Official: https://www.anthropic.com/claude-opus-5-5. Docs: https://platform.claude.com/docs/en/models/opus-5-5/overview.

confirmed · 2026-09-21 · L2–L4 workflow

Grok 4.7 is available — coding / knowledge-work model; Cursor + Grok Build + API — not a free course

Monday: xAI shipped Grok 4.7 (Cursor blog Sep 21, 2026; official https://x.ai/news/grok-4-7). Larger base than Grok 4.6; longer RL on multi-hour tasks; same $2/$6 per MTok list price as 4.6, with a faster variant at 2× speed / 2× price. Available today in Cursor, Grok Build, Grok API, third-party coding harnesses and routers. This is a paid model / IDE surface, not a free course — it does not replace any catalog door. Aug 26 SuperGrok Bot gate item still true for Grok Bot product access; this WIRE is the model GA. Do not invent an xAI Academy door. Official: https://x.ai/news/grok-4-7. Cursor: https://cursor.com/blog/grok-4-7.

confirmed · 2026-09-22 · L4 radar

Xiaomi open-sourced MiMo-V2.6-Pro / Flash — strongest open-source claim vs Fable 5.1 / Astra gap

Tuesday: Xiaomi MiMo released and open-sourced MiMo-V2.6-Pro and MiMo-V2.6-Flash (official update Sep 22, 2026). Native multimodal; Pro claims AA Intelligence Index 46 and top open-source ranking while still trailing Claude Fable 5.1 and GPT-6 Astra. Weights + report on Hugging Face (XiaomiMiMo/mimo-v26 collection); API on Xiaomi MiMo Open Platform at prior V2.5 prices. L4 open-weights / HF paths stay valid; do not invent a new free academy door; do not catalog Mythos-style cyber variants as PATH. Official: https://mimo.mi.com/docs/en-US/news/latest/v2-6. HF: https://huggingface.co/collections/XiaomiMiMo/mimo-v26.

confirmed · 2026-09-16 · L1–L4 workflow

Gemini 3.8 Live / Live Extended Thinking rolling out — Search Live for everyone; not a free course

Tuesday: Google’s official blog (Sep 15, 2026) introduced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking — voice / live-dialogue models. Official access: Live rolling out to developers (Gemini API / AI Studio), enterprises (private preview), and everyone in Search Live. Extended Thinking rolling out to API / AI Studio, Enterprise private preview, everyone in Gemini Live, Google AI Pro/Ultra in Workspace Docs, and all Google AI subscribers in Gmail and Keep. Dev Live API list price $0.005/min audio in / $0.018/min audio out (developer blog). This is a model surface, not a free course — it does not replace Grow with Google: https://grow.google/ai/. Flash 3.8 paid-gate story unchanged. Flash Cyber / Fairwind stay off PATH. Do not catalog. Official: https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-8-live-gemini-3-8-live-extended-thinking/. Dev audio: https://blog.google/innovation-and-ai/technology/developers-tools/build-real-time-voice-applications-gemini-audio/.

confirmed · 2026-09-14 · L2–L4 workflow

Claude Code weekly limits settled: promo ended; permanent +25% vs pre-May baseline

Monday: Anthropic Help Center restamped the Claude Code weekly-limits promotion (Updated today). The temporary +50% weekly boost ran May 13 through September 13, 2026 (11:59 PM PT) for Pro / Max / Team / legacy seat-based Enterprise — Free and consumption-based Enterprise were never in it. Starting September 14, 2026, weekly limits in Claude Code are permanently 25% higher than before the promotion (Help Center wording). Vs the promotional level that just ended, that is a step down (~17% vs yesterday’s boosted bar per Anthropic’s earlier ClaudeDevs arithmetic). 5-hour limits were never part of this promo. Claude Code only (CLI / IDE / desktop / web); Claude chat and Cowork weekly limits unchanged. Check Settings → Usage on claude.ai or run /usage in Claude Code — do not plan from secondary recaps alone. This is a usage-limit move on a paid tooling surface, not a free course. Free door stays Anthropic Academy: https://anthropic.skilljar.com/. Official: https://support.claude.com/en/articles/15910845-claude-code-may-august-2026-weekly-limits-promotion. Fable/Mythos plan rules unchanged: https://support.claude.com/en/articles/15424964-claude-fable-5-on-your-plan.

confirmed · 2026-09-10 · L2–L4 workflow

GPT-6 Astra is available on paid ChatGPT/API — not a free course

Wednesday: OpenAI Help Center flipped GPT-6 Pro (Astra) from rolling out to available for Pro $100 / Pro $200 / Business / Enterprise (Enterprise still needs workspace model-access permission). Plus gets Astra in ChatGPT Work and Codex, not Chat GPT-6 Pro. Chat allowances are not unlimited — Pro $200: 200/wk; Pro $100: 50/wk shared with Sol Pro; Business Standard: 15/mo; Business Premium: 50/wk. Work/Codex allowances are separate. This is still a model, not a free course — it does not replace OpenAI Academy: https://academy.openai.com/. Daybreak / critical cyber threshold is not a PATH door. Do not catalog. Do not send working adults to apply. Live availability: https://help.openai.com/en/articles/20001354-gpt-5-6-in-chatgpt. Launch post (still uses early rolling-out wording): https://openai.com/index/gpt-6-astra/. Path/safety: https://openai.com/index/path-to-astra/

confirmed · 2026-09-03 · L4 radar

NVIDIA agreed to buy Hugging Face (about $12.9 billion) — not closed yet; hub stays open

Correction to our Aug 27 “reported” item: NVIDIA announced a definitive agreement to acquire Hugging Face (official NVIDIA blog, Sep 3, 2026). NVIDIA says Hugging Face stays an open platform and NVIDIA compute will not be required to build on or deploy through it. The deal still needs regulatory approval (NVIDIA 8-K: expected close in the first half of 2027), so it has not closed. L4 paths that send people to huggingface.co/learn stay valid today. Official: https://blogs.nvidia.com/blog/nvidia-to-acquire-hugging-face/.

confirmed · 2026-09-02 · L2–L4 workflow

Gemini 3.8 Flash is generally available. Paid gate: Google AI Pro/Ultra or API / AI Studio — not a free course

Wednesday: Google shipped Gemini 3.8 Flash. This is a model, not a free course — it does not replace Grow with Google: https://grow.google/ai/. Consumers: Google AI Pro/Ultra (Gemini app). Devs: API / AI Studio. Intro API $0.75/$3.75 per MTok through Dec 31, 2026. Same family as Fable: note the paid gate, not a new free door. Flash Cyber is Fairwind trusted-defender only — not a PATH door. Do not catalog. Do not send working adults to apply. Official: https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/

confirmed · 2026-09-01 · L2–L4 workflow

Claude Fable 5.1 is generally available. Anthropic reset 5-hour and weekly usage — check Settings → Usage

Monday: on claude.ai open Settings → Usage. Anthropic reset 5-hour and weekly counters for all users with this launch — not the July reset. Fable 5.1 is not on Free. Pro / standard Team seats do not include it in plan limits (usage credits or upgrade). Max / premium Team / premium Enterprise include Fable up to 50% of weekly limits and it burns faster. The July 19, 2026 Fable 5 promotion is over. Same family as Fable 5; start most work on Opus 5, use Fable 5.1 for long-horizon agentic work or when Opus 5 is not enough. Claude Code’s default Fable is now 5.1. Official: https://www.anthropic.com/claude-fable-and-mythos-5-1. Plan rules: https://support.claude.com/en/articles/15424964-claude-fable-5-on-your-plan. Free course door: https://anthropic.skilljar.com/. Mythos 5.1 is trusted access only (cyber + life sciences) — not a PATH door.

confirmed · 2026-08-26 · L2–L4 workflow

Grok Bot is on SuperGrok ($30/mo), not only Heavy

In two weeks the gate moved Heavy → Plus → SuperGrok. If the office default was “bots are a $300 toy,” that sentence is already stale. Human gate before it sends mail or touches a client file.

confirmed · 2026-08-27 · Career

Microsoft AI-900 already retired (June 30). AI-901 is the live fundamentals exam (~$99)

Old study plans still say AI-900. Career track: free badges first; pay $99 only when the résumé line needs the current Microsoft code.

confirmed · 2026-08-23 · L0–L1

Google door is Grow with Google: AI training at grow.google/ai/

Official free Google on-ramp. Modules that stay free on grow.google. Cloud Skills Boost intros and the Machine Learning Crash Course stay linked.