I wanted a Blender roadmap with the clarity of Andrej Karpathy’s Zero to Hero.
One skill. One build. Something you can actually finish.
So I put together 12 skills to practice, with open source repos to explore:
Exploded assemblies
Build a camera with separated optical groups.
https://t.co/kHKOafR6Ra
Layered modeling
Build a sneaker from outsole to knitted upper.
https://t.co/9YjQkqgoyf
Mechanical rigging
Make a robotic hand articulate.
https://t.co/YPNSTvNfQR
Curves and subdivision
Shape a bentwood lounge chair.
https://t.co/mlxmoo7Nnl
Hard surface modeling
Build an FPV drone with detailed components.
https://t.co/1Tjt1MtwO3
Cloth simulation
Make fabric drape over furniture.
https://t.co/YPNSTvNfQR
Precision modeling
Build the components of a mechanical watch.
https://t.co/1Tjt1MtwO3
Modeling internals
Reveal the plumbing inside an espresso machine.
https://t.co/mlxmoo7Nnl
Constraints and pivots
Animate an articulated desk lamp.
https://t.co/YPNSTvNfQR
Material layering
Combine leather, metal and mesh in headphones.
https://t.co/YPNSTvNfQR
Automotive modeling
Practice bodywork, surface continuity and reflections.
https://t.co/YPNSTvNfQR
Arrays and repetition
Build a keyboard from switches, keycaps and layers.
https://t.co/kHKOafR6Ra
These repos contain tools and source code to explore. They are not tutorials for the exact objects shown.
Pick one object. Finish it. Let the build reveal what you need to learn next.
Save this for your next Blender project ↓
Paste this into Claude Fable 5.2 tonight, GTA 6 will change your life...
it will build a GTA 6 clipping page that can make up to $15k/month with prompts, workflows, and everything for you
Steal this before everyone starts using it:
Paste this into Claude Fable 5.2 tonight, it will change your life...
it will build an entire AI team to handle it with AI agents, prompts, and workflows for you
Save it before your timeline catches on:
Anthropic daje za darmo cały kurs o tym, jak dobrze pracować z AI. 14 lekcji, ok. 4 godziny, a na koniec test i certyfikat.
Zrobili go razem z dwoma profesorami: Josephem Fellerem z University College Cork i Rickiem Dakanem z Ringling College. Jest dla ludzi, którzy używają Claude albo ChatGPT w pracy, programować nie trzeba.
Cały kurs stoi na 4 umiejętnościach:
- Delegowanie, czyli co oddać AI, a co zrobić samemu, zanim w ogóle otworzysz czat
- Opis, czyli jak powiedzieć AI, czego chcesz, żeby nie musiało zgadywać
- Ocena tego, co dostajesz, zanim uznasz to za dobre i prawdziwe
- Staranność, czyli za co odpowiadasz Ty, gdy wysyłasz dalej coś zrobionego z AI
Do tego osobna lekcja o tym, czym jest generatywna AI i czego nie potrafi, oraz lekcja z technikami pisania promptów.
Kurs jest po angielsku. Jeśli masz w tym tygodniu jeden wolny wieczór, zacznij od lekcji o delegowaniu, bo od niej zależy cała reszta.
the #1 model from Jane Street's contest is free on GitHub... & there's an AI hedge fund JPMorgan reserved $500M for that pays for models like it
Jane Street threw its own trading data at 4,245 teams to find the best model... this is the one that WON
Numerai is a hedge fund that trades the combined predictions of thousands of outside models. JPMorgan reserved up to $500M in it, it runs about $700M, & it pays NMR to models that stake & score
first understand how this works:
> both give you anonymized market data & ask one question: which way next
> the winning model: an autoencoder that strips the noise, an MLP on top & XGBoost next to it, validated so it never peeks at the future
> Numerai scores your predictions on live stocks every round & the fund trades the combined signal. stake NMR & good scores pay, bad ones burn
how to run it with Claude Code:
1. tell Claude Code to clone https://t.co/Jszd7YIYkt & numerai/example-scripts
2. open hello_numerai: it pulls Numerai's free data (about 8 GB) & trains the baseline
3. ask Claude to port the winning AE-MLP + XGBoost to Numerai's features & targets, swapping the GPU-only parts (cudf, cupy) for pandas
4. keep the winner's purged time-split, so validation never sees the future
5. score both models on Numerai's validation data: correlation & Sharpe per era
6. ask Claude to connect Numerai's MCP server, it's one command from their docs
7. make a free account & upload your predictions every round
8. ask Claude to compare your live scores with validation every week
9. stake NMR only when the live scores hold up
i won't say anything more:
YANN LECUN: "A STUDENT SHOULD NOT TRY TO MAKE THEMSELVES USEFUL"
if you had the choice between a course on iOS programming or quantum mechanics, "take quantum mechanics"
In 40 seconds he explains:
→ why tech courses are "very likely to be out of date 5 years from now"
→ why your profession is "almost certainly" going to change during your career
→ why the only skill that doesn't expire is learning to learn
the World Economic Forum expects 39% of workers' core skills to change by 2030
lecun won the Turing Award for the neural networks behind today's AI, and he still tells students to learn the stuff that won't expire
Free fundamentals he'd sign off on:
→ The Feynman Lectures on Physics, all 3 volumes free to read online. Feynman is one of lecun's scientific heroes
https://t.co/2yAvSyVTho
→ MIT's full Quantum Physics I course, free lecture videos
https://t.co/HFxQkFPQ5U
→ 3Blue1Brown's Essence of Linear Algebra, the math every neural network runs on
https://t.co/Zqcxrp7uzM
→ Learning How to Learn, 4.2M+ students, co-taught by Terry Sejnowski, who built early neural networks with Hinton
https://t.co/iEwojpsiYz
the course that looks most useful today usually has the shortest shelf life
and if you want to see how the people building AI actually learn and work with it, I broke it down in my article ↓
La mayoria de las maquinas se descomponen y nadie sabe la razón
Alguien modeló en Blender, todo con código, una cortadora de acero.
Cómo funciona la demo:
→ La barra gira en el chuck y la herramienta avanza quieta a lo largo
→ Cada vuelta le quita una tira de acero
→ Tres pasadas de 1.5 mm y una fina la bajan de 40 a 30.5 mm
→ En vivo ves las revoluciones, la fuerza de corte y la potencia
Para qué sirve:
→ Entender cómo se fabrican ejes, tornillos y piezas redondas
→ Aprender las partes de un torno antes de usar uno real
this is f*cking close. the two strongest AI models alive made the same bear - and they came out nearly identical.
Claude Fable 5.2 · GPT-6.2 Astra
no edits, no favorites. just both, side by side.
here’s the kicker: one of them costs way more to run than the other. so would you pay up for it, or take the cheap one?
drop your pick.
Si siempre quisiste convertir cualquier industria en un simulador en vivo, ve esto.
Alguien hizo una skill gratis para Claude que arma simuladores 3D, y con ella levantó una obra completa: de la tierra pelona a los muros.
Cómo funciona la demo:
→ Las excavadoras cavan el foso y los camiones se llevan la tierra
→ Llega la revolvedora, cuela la losa y el avance sube en vivo
→ En otras dos casas ves los muros de block subiendo
→ Le das clic a cada máquina y ves qué está haciendo
Para qué sirve:
→ Armar demos de cualquier operación sin modelar todo a mano
→ Ver cómo Claude Code reparte el trabajo: 2 subagentes y 3,829 líneas
this is f**king dangerous
watched an 18-year-old kid casually show how he lets Claude Code play GTA 6 all night while he sleeps, and it made him $11,860
the entire setup:
give Claude Code access to your screen. type one rule: "play GTA 6 all night. if you die, i cancel your $200 subscription"
from that point on, AI does everything plays, farms money, records, cuts in CapCut, captions, posts to TikTok, Twitch, YouTube.
> 12 hours of gameplay ≈ 40 min of clips
> 10M views ≈ $11,860 in payouts
> zero code, zero editing, zero hours played
you're not playing the game anymore. the game is paying you.
the full prompt is in the article below exact prompt, exact tools, exact setup
save it
Microsoft quietly dropped one of the CLEANEST FREE AI agent courses on GitHub
Building AI agents is the skill companies are hiring for right now
18 lessons. Real code. Design patterns. Tool use. Agentic RAG. Multi-agent systems. Memory. Context engineering. Production deployment. Security
It’s the practical path most people are still paying $2,000+ for
01 Intro to AI Agents and use cases
02 Exploring agentic frameworks
03 Agentic design patterns
04 Tool use design pattern
05 Agentic RAG
06 Building trustworthy AI agents
07 Planning design pattern
08 Multi-agent design pattern
09 Metacognition design pattern
10 AI agents in production
11 Agentic protocols (MCP, A2A)
12 Context engineering
13 Agent memory management
14 Multi-agent collaboration
15 Computer-use agents
16 Scaling and deployment
17 Local AI agents
18 Securing AI agents
Each lesson has written content + Python code samples + short video
Microsoft put the full thing on GitHub for free
80k+ stars already
You either start building real agents now, or keep watching random YouTube videos for another year
Save this ↓
this harness paper is f*cking gold
Anthropic engineers laid out what actually makes Claude Code work so well. spoiler: it's not the model.
the secret: the harness. that's the entire gap.
prompt engineering → context engineering → harness engineering.
the 6 layers of the harness:
layer 1 → tools
fewer tools, one job each.
every error tells the agent what to try next.
layer 2 → context
short claude-md → files on demand → compaction when it fills up.
layer 3 → permissions
reads + tests → auto.
deploys + deletes → ask a human.
everything → sandbox.
layer 4 → hooks
before a tool runs → block secrets and protected files.
after every edit → format, lint, typecheck. no exceptions.
layer 5 → verification
fresh reviewer reads the diff, not the summary → runs the tests → fail? fix → rerun.
layer 6 → evals
every harness change scored → success, tokens, turns → keep what wins.
the model is the engine. the harness is the whole car.
they even packed all 6 layers into one prompt. paste it into Claude Code and it engineers the harness for you.
6 layers. 1 prompt. 0 infinite loops.
then read the article below ↓
Alguien acaba de publicar el manual de instrucciones del fin del mundo.
Se llama Lastbook. Un programa gratuito que guarda en tu ordenador cómo potabilizar agua, hacer conservas, orientarte con mapa y brújula y qué hacer ante una herida que no deja de sangrar.
Y puedes añadirle una IA que te responde sin internet.
Código abierto. Sin cuenta.
El día que no puedas buscarlo, ya lo tendrás.
holy f*ck, this is smart - Anthropic's GitHub is basically a full AI degree
$60,000+ at a university
$0 here: 9 repos, 317,602 stars
the final exam is in there too, the hiring test Claude started beating most candidates at
the curriculum:
claude-code (153,451) - Claude Code itself
↳ https://t.co/qW9YECQdee
claude-plugins-official (44,712) - official Claude Code plugins
↳ https://t.co/TU9y47rQHl
financial-services (40,800) - agents and skills for banking and research
↳ https://t.co/oXH9yIWMcV
knowledge-work-plugins (33,170) - Claude as a specialist for your role
↳ https://t.co/8U1NsCqqy9
claude-quickstarts (19,003) - starter projects to deploy
↳ https://t.co/lNPqXZjh7Y
claude-for-legal (9,795) - plugins for legal workflows
↳ https://t.co/18xoVxoK6S
claude-code-action (9,651) - Claude Code inside your PRs
↳ https://t.co/zRfpXljzQv
original_performance_takehome (4,198) - the retired hiring test, open to try
↳ https://t.co/0v0990wF8E
claudes-c-compiler (2,822) - a C compiler Claude wrote in Rust
↳ https://t.co/rkvnufoqAk
try the take-home before you apply anywhere in AI
it shows exactly where the bar sits
the Claude skills worth adding on top, all 32 of them, are below ↓