And here we go…GPT-6.2 Astra and GPT-6.2 Sol are yours to make more games with.
Oh, and they’re also super efficient and really good at making slides, building websites, and tackling the work you were doing before you started making games.
Rolling out today to ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users.
https://t.co/tjQBDQWB2n
It's f*cking insane that you can now train your own ChatGPT-style model from zero in 2 hours for less than a dollar
it's called MiniMind (65.8K stars), and it walks you through every step big labs take, on a model 2,700x smaller than GPT-3:
→ tokenizer: turns text into numbers
→ pretraining: learns language by guessing the next word
→ fine-tuning: learns to answer like an assistant
→ DPO and RL: learns which answers are better
→ tool use: learns to call a calculator or a search
→ chat: runs on your laptop or behind an OpenAI-style API
no black-box libraries. every algorithm is written in plain PyTorch, so you can actually read how it works
one GPU, 2 hours, $0.40. that's it
https://t.co/9p0YhY3oai
save it for your weekend build
Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me.
I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA.
I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation.
Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO.
Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours.
Jensen
THIS IS AN ACTUAL F*CKING GOLDMINE
25 best skills for Fable 5.2 & Opus 5.5
build websites. animate videos. write copy. audit SEO. test your code
Fable for planning → Opus for execution
DESIGN IT → PACKAGE IT → MARKET IT → BUILD IT → VERIFY IT
the full collection ↓
01 / DESIGN + MOTION
1. Frontend Design
Build polished websites and interfaces.
https://t.co/HG3a1tv6Vw
2. Canvas Design
Create posters, compositions + visual assets.
https://t.co/1ifKx5SgIj
3. Algorithmic Art
Generate art with code.
https://t.co/yiINocXVLa
4. Theme Factory
Apply consistent colors + typography to your work.
https://t.co/tV0TVR5Haz
5. Remotion Create
Build animated videos with React.
https://t.co/2Kt21l3hUw
02 / DOCUMENTS + CONTENT
6. DOCX
Create and edit Word documents.
https://t.co/6Sm0hcCM8c
7. PPTX
Build presentations and slide decks.
https://t.co/utfk5MK4ld
8. XLSX
Work with spreadsheets, formulas + charts.
https://t.co/kR8mlmkrMZ
9. PDF
Extract, combine and process PDFs.
https://t.co/JUkrkikCwV
10. Doc Coauthoring
Turn rough ideas into structured documents.
https://t.co/OxFYEq8EwE
03 / MARKETING + GROWTH
11. Copywriting
Write landing pages, headlines + CTAs.
https://t.co/88j2mfXIO2
12. SEO Audit
Find technical and on-page SEO issues.
https://t.co/DPfr22FpAR
13. CRO
Review pages and forms for conversion friction.
https://t.co/ZcELZVs4O3
14. Emails
Create email campaigns and sequences.
https://t.co/JTbK7mT3YW
15. Social
Plan and write social content.
https://t.co/ICFpwge6TZ
04 / DEVELOPMENT + TOOLS
16. Writing Plans
Break a feature into actionable implementation steps.
https://t.co/VdnhU4DyTy
17. Subagent-Driven Development
Delegate implementation tasks with review steps.
https://t.co/MnBwTiaVpz
18. Git Worktrees
Keep parallel work in isolated workspaces.
https://t.co/T56GvTciMw
19. MCP Builder
Build MCP servers for tools and APIs.
https://t.co/kylg8FH5UY
20. Skill Creator
Create, evaluate and improve your own skills.
https://t.co/TVS3NpE9VX
05 / TESTING + VERIFICATION
21. Webapp Testing
Test browser interactions with Playwright.
https://t.co/j2irB1UmJr
22. Systematic Debugging
Trace bugs to their root cause.
https://t.co/l1doObdpTM
23. Test-Driven Development
Write the test → implement → refactor.
https://t.co/b7uiN5VlVE
24. Requesting Code Review
Get a structured review of your changes.
https://t.co/yuvYf7hY72
25. Verification Before Completion
Check the evidence before calling a task done.
https://t.co/gJobLc4R7Q
save this for your next build → pick the skills you need
then build an AI team with Fable 5.2 & an AI employee for your business ↓
This is f*cking next level, NVIDIA, Google, Meta and Tencent all have new agent repos trending on GitHub, and these 8 are the ones to run with Fable 5.2
-> skillspector (by NVIDIA) https://t.co/k1tzZNy9AT
scans a skill for vulnerabilities before you install it
-> elephant goldfish model (by google) https://t.co/XMm00Qwr4i
one agent writes the spec, fresh agents review it with clean eyes
-> t3code (by pingdotgg) https://t.co/ckfq8vAPEC
run Claude Code and other coding agents from your phone
-> plannotator (by backnotprop) https://t.co/JebAl4Rx2D
mark up the agent’s plan in your browser and send your notes back
-> billion context (by ranxianglei) https://t.co/N4MWSUyEQ2
keep one Claude Code session alive for weeks by compressing the conversation
-> loopforge (by Tencent) https://t.co/z9cHd4bU47
run coding agents through repeatable SOPs, with sandboxes, approval gates and checkpoints
-> agent desktop (by vercel-labs) https://t.co/bce5WPs45V
lets an agent operate desktop apps through their accessibility layer
-> astryx (by meta) https://t.co/P6N3fdIUmb
a design system of 150+ interface parts, documented for agents to use
Fable 5.2 does the work
these repos keep it safe, planned and running
i genuinely can't wrap my head around why this isn't everywhere yet. it's the one setup that turns everything you've ever read into something your AI actually remembers
Andrej Karpathy, one of OpenAI's founders, wrote a short idea file built to be fed to an agent. the idea: stop asking your AI for answers, let it build a library of everything you've ever studied
you drop a source into a folder. Claude Code reads it, ties it to every page it already wrote, and keeps growing the library. each new file quietly upgrades the old ones
how it's built:
→ one folder, a raw subfolder for your sources, Fable as the model
→ start with a handful, 3 to 10 files
→ paste the idea file and say: let's build this together
→ your sources stay as they are, Claude keeps its own pages with an index and a change log
→ house rules go into a CLAUDE file, so tomorrow's session picks up exactly where today's stopped
it's working when it can answer 5 questions that no single source in the pile could
from then on, nothing you read is wasted
I used Claude Fable 5.1 to build a self-learning trading algorithm and made +$3,275.30 in 3 days
1,268 token launches tracked on pumpfun. 25,360 holders analyzed. 67 wallet clusters detected. Only 4 tokens passed every check, each scoring 90+/100
I MADE THIS ALGORITHM PUBLICLY AVAILABLE
GitHub: https://t.co/my5H653LfH
It identifies tokens with high growth potential by examining their developers, holders and conducting a full on-chain analysis to uncover potential risks
A recent case:
Oct 25, 13:27 UTC: mihao launched on pumpfun
Nov 9, 14:45 UTC: Market cap reached $580K → QuantumOrb scored it 92/100
2 days later, its market cap reached $15.2M (26.2x after the alert)
Top 20 holders → 20 operators → 0 clusters
Estimated selling impact → −18.7%
Here’s what happens inside the algorithm:
Five components handle different stages of the analysis
> RADAR detects new pumpfun launches and monitors changes in market cap, trading activity and bonding curve progress
> SHIELD examines token creators, identifying repeated unsuccessful launches and unusually large developer purchases
> SCANNER reconstructs token distribution by analyzing the top 20 holders and first 20 buyers
For example, the algorithm measures holder concentration using:
Top20Share = Σ(HolderShareᵢ)
ConcentrationScore = 10 × clamp((0.70 - Top20Share) / 0.45, 0, 1)
> CARTOGRAPHER traces transaction connections, identifying wallets with shared token sources or direct on-chain links
> GUARD evaluates remaining developer holdings, launch-slot accumulation and previous token launches
For example, early token accumulation and transferred supply are analyzed using:
SniperShare = Σ(UnsoldSniperShareᵢ)
SniperScore = 10 × clamp((0.15 - SniperShare) / 0.14, 0, 1)
TransferShare = Σ(TransferredHolderShareᵢ)
TransferScore = 20 × clamp((0.10 - TransferShare) / 0.09, 0, 1)
The final result is a risk score from 0 to 100. Critical red flags can override the verdict regardless of the numerical score
A chart shows how far a token has already moved
QuantumOrb investigates who controls its supply, what risks may be hiding behind that growth and what could happen to its price next
THIS IS A REAL F*CKING GOLDMINE. a year ago the best AI finished 2.5% of real freelance jobs. Fable 5.2 finishes 41.3%. this board runs the team module from this article: 1 Fable brain, 1,000+ Opus 5.5 workers, 52,000 tasks.
stop letting one model plan the job, type it and grade it.
how the run splits:
→ Fable 5.2 plans every task at high effort and hands it to its own Opus 5.5 worker
→ the worker types it at half the price, with "you own this project" in its prompt. that line gets the work cleaned up in ~75% of runs, under 10% without it
→ Fable checks the result: run it, screenshot it, fix what looks wrong
→ a pass ships into its build: game, app, site, film, team, music, business, device. a fail goes back to the worker with the reason
You only look at three moments:
→ the finish line. on this concept board "no finish line" is the top reason a task ends as a demo, 2,713 of 7,140. the model won't write it for you
→ the check. 1,571 more ended on "check never ran". add "do not report a check as passed unless you ran it" and read what it actually ran
→ the decision. the conductor only stops when it needs you. that stop is the one message you don't skim
the first line of my prompt on runs like this: "Show every diff first. Do not apply edits until I say go."
the part worth copying is the split: Fable thinks and checks, Opus types.
on your last Fable run, which did you skip: the finish line, the check or the stop for a decision?
This is f*cking unreal, 20 free repos that build your whole second brain from capture to AI memory, and people still start with a blank page
(1) CAPTURE: get everything in
-> markitdown (by microsoft) https://t.co/zrR2oFNAMI
turns PDFs, Office files, images, audio and web pages into Markdown
-> defuddle (by kepano) https://t.co/rhYFFYsYRE
clean Markdown from any web page
-> whisper (by openai) https://t.co/25kgHJOWRA
turns audio into text
-> grimoire (by goniszewski) https://t.co/TfdbQ7Ofv0
bookmarks with semantic search
(2) STORE: where it lives
-> logseq (by logseq) https://t.co/2fH4iKiIbU
open source outliner and personal wiki
-> siyuan (by siyuan-note) https://t.co/gJxB4HjanU
a self-hosted workspace where you and AI agents work together
-> joplin (by laurent22) https://t.co/t2h4zRGCPN
private notes that sync across your devices
-> memos (by usememos) https://t.co/DPxide7Q4E
quick daily notes on a personal timeline
(3) STRUCTURE: how it is organized
-> kepano obsidian (by kepano) https://t.co/vflEVjrbF4
the Obsidian CEO’s own vault, steal the structure
-> obsidian templates (by groepl) https://t.co/y4J1LDxqF1
Zettelkasten templates and scripts
-> digital gardeners (by MaggieAppleton) https://t.co/NRZ920SNs6
ideas for growing notes that last
-> jsoncanvas (by obsidianmd) https://t.co/qUkoioW7He
an open canvas format, your maps stay plain files
(4) AI: let the agent maintain it
-> claude obsidian (by AgriciDaniel) https://t.co/RNFghwuqU5
Claude files every source into a linked Markdown graph
-> obsidian mind (by breferrari) https://t.co/0Z3MLY54ct
a vault structure that gives coding agents persistent memory
-> llm wiki agent (by SamurAIGPT) https://t.co/UA9arBaHlR
an agent reads your sources and builds a wiki that maintains itself
-> obsidian skills (by kepano) https://t.co/lonWXdfFdE
skills that teach Claude to read and write your vault correctly
(5) MEMORY: find it again
-> OB1 (by NateBJones-Projects) https://t.co/rM42kCL2VU
one memory layer any AI tool can connect to
-> khoj (by khoj-ai) https://t.co/bccxuKqWNK
a self-hosted AI that answers from your own notes
-> anything llm (by Mintplex-Labs) https://t.co/QQeIXD0J5N
chat with your documents using many models
-> graphify (by safishamsi) https://t.co/W4WmD7ZTrP
a knowledge graph from code, docs, papers and images
capture, store, structure, think, remember
that is the whole second brain, and every piece is free
It's f*cking scary how fast way too many plugins turn your second brain into a mess while the CEO of Obsidian keeps his clean with 3 rules
the folders are the easy part. the real system is the filter: what's allowed in, and what gets thrown out
a perfect template won't save a messy inbox. it just gives the mess better labels
and every plugin you add is one more thing that breaks after an update
3 of his rules that keep it clean:
→ almost no folders, so nothing gets lost three levels deep
→ quick notes get a timestamp and a review every few days
→ random revisits: he reopens old notes and fixes them up
https://t.co/2r4QoZ7DLN
the filter, in 6 repos:
Find-unlinked-files (240K downloads)
finds notes nothing links to and broken links, then deletes or archives them
https://t.co/lRnmI4ISbs
Obsidian-janitor (65K downloads) give a note an expiry date and it cleans itself up, plus empty files and dead attachments https://t.co/WqfxEQwVDw
Auto-note-mover (127K downloads)
tag a note once, it moves itself out of the inbox
https://t.co/iy7rSyBIpP
Dataview (9.4K stars)
one query shows every note you haven't touched in 90 days
https://t.co/yaV3XI5JLM
Obsidian git (12.2K stars)
full history of your vault, so deleting stops being scary
https://t.co/3mj3xzsxEK
Obsidian skills (48.5K stars)
by Obsidian's CEO. Claude Code goes through your inbox: keep, merge or delete
https://t.co/SG7k1XpYrK
copy his skeleton. build your own filter
save it so you don't lose it
This is f*cking insane.. Anthropic just dropped the best trick for running a long-running agent harness
Fable 5.2 advises → Sonnet 5.5 builds one feature → Opus 5.5 tests it in the browser → Haiku 5.5 swarms the files → git commit + progress note → next session picks up where the last one stopped
your agent forgets everything every session. the files don't.
prompt + Anthropic's official docs below ↓
Anthropic’s GitHub is a rabbit hole.
I went looking for one repo to build a Claude agent.
I ended up mapping 12: agent SDKs, tool connections, reusable skills, GitHub workflows, example apps and courses.
Here’s the map I wish I’d found first ↓
RUN THE AGENT → CONNECT TOOLS → BUILD WORKFLOWS → LEARN THE STACK
01 / RUN THE AGENT
1. Claude Agent SDK — Python
Build agent applications in Python.
https://t.co/xguka3Giwv
2. Claude Agent SDK — TypeScript
Build agent applications in TypeScript.
https://t.co/LNTiuifzhH
3. Claude Code
Explore the terminal coding agent, documentation and examples.
https://t.co/SkjKYCjDnK
02 / CONNECT + EXTEND
4. MCP — Python SDK
Build MCP servers and clients in Python.
https://t.co/IzGJd3RMcI
5. MCP — TypeScript SDK
Build MCP servers and clients in TypeScript.
https://t.co/sNtQ2h1Qol
6. Agent Skills
Give Claude reusable instructions and resources for specific tasks.
https://t.co/92AxiUAo9k
03 / WORKFLOWS + EXAMPLES
7. Claude Code Action
Bring Claude Code into GitHub Actions workflows.
https://t.co/YtEj3OJm3L
8. Claude Quickstarts
Start from example applications built with the Claude API.
https://t.co/zIMrzIXuqe
9. Anthropic Cookbook
Explore notebooks and recipes for practical Claude use cases.
https://t.co/OrWYzPRA3H
04 / API + LEARNING
10. Anthropic Python SDK
Call the Claude API from Python applications.
https://t.co/EBU0sO5xAA
11. Anthropic TypeScript SDK
Call the Claude API from TypeScript applications.
https://t.co/mtexhy7hTO
12. Anthropic Courses
Work through Anthropic’s educational materials.
https://t.co/vxB3U9odtt
My starting route:
Agent SDK → one useful tool → one working task → inspect the result.
Add skills and automation when the task calls for them.
The repos are public. Running the applications may require API access and paid usage.
Save this before your next Claude agent build ↓
your Claude forgets everything the moment the session ends. 7 open-source memory layers fix that, 310K+ stars
1. claude-mem · 117.5K ⭐
captures what your agent does in a session, compresses it and injects it into the next one
https://t.co/i1ICMFTSFk
2. mem0 · 69K ⭐
a drop-in memory layer for agents and apps. context that persists
https://t.co/tCr28sOLBm
3. cognee · 33.3K ⭐
open-source long-term memory platform for agents
https://t.co/7TrpRqoTja
4. graphiti · 32.8K ⭐
real-time knowledge graphs that track how facts change instead of trusting the first note
https://t.co/7DMdrXiWBn
5. supermemory · 31.6K ⭐
fast memory and context engine that can run fully local
https://t.co/GQzQPIcopI
6. letta · 25.5K ⭐
stateful agents with memory that learn and improve over time
https://t.co/axYi6coOKh
7. basic-memory · 4.6K ⭐
plain markdown memory your AI reads and writes. never re-explain your project
https://t.co/5knUr3CAjT
start tonight, inside Claude Code:
/plugin marketplace add thedotmack/claude-mem
/plugin install claude-mem
next session starts with context instead of a blank page
the guide below shows how to make that memory prove what it knows
WHOEVER PUT THIS ON GITHUB IS F*CKING INSANE.
someone just dumped 300+ AI agents into a single GitHub repo.
and holy sh*t, it’s basically a cheat sheet for which ones are actually worth using.
here's what I'd plug into your LLM:
want the agent loop without writing it from scratch?
-> Claude Agent SDK
want docs in its context that aren't from 2023?
-> Context7
want memory that knows what changed and when?
-> Graphiti
want bash running in a box, not on your laptop?
-> sandbox-runtime
want to catch a sketchy MCP server before your agent trusts it?
-> MCP Scanner
want to see every tool call after it does something weird?
-> Langfuse
want to break your agent before someone else does?
-> promptfoo / garak
stop swapping models every week. fix the harness.
save this for your next agent build.
Forget every other list, these 10 GitHub repos are all your Fable 5.2 will ever need
only the f*cking best for Fable 5.2
Polish. Motion. 3D. Visuals. Captions. Export.
Impeccable → turns "AI-looking" into "who designed this?"
https://t.co/LUVwZwOvKe
GSAP Skills → teaches your agent smooth animation, straight from GSAP's team
https://t.co/dTUBVoOjKQ
Three.js → 3D scenes that play right in your browser
https://t.co/BgbvnRcMUt
Blender MCP → your agent gets hands inside Blender
https://t.co/QWCqTa15R2
anatomy → shows how any machine works by taking it apart
https://t.co/h9AZjPTBp2
ComfyUI → makes AI images and video on your own computer
https://t.co/RN6kGEDwOT
faster-whisper → subtitles from any voice, way faster than Whisper
https://t.co/yiMYGYEzYs
Remotion → builds whole videos out of code
https://t.co/znpMp7SpNi
HyperFrames → HeyGen's free tool that turns web pages into videos
https://t.co/dOZZVJZopQ
FFmpeg → exports the final animation
https://t.co/HlJpuzR3nn
Polish → motion → 3D → visuals → captions → export
Lock this in, then go make the best thing you've ever made with Fable 5.2 and the guide below ↓
Mio’s Pose Designs Vol. 27
白いマネキンモデルを使って、友達とのお出かけや記念撮影をテーマに、9種類のポーズを制作しました✨
引用ポストも、ポーズの参考にするのもご自由に!
友達と一緒に一番やってみたいことは何ですか���
I created nine different poses for outings and photos with friends using white mannequin models as references for artistic creations! ✨
Feel free to QT and use these poses as references!
Which activity would you most like to do with a friend?