tengyuan

[Vikas gupta @vicky_grok]

Anthropic pays $750,000+ a year for engineers who can build LLM architectures from scratch. Stanford taught the entire thing in 1 hour lecture & released it for free.

Bookmark & watch this today before someone takes it down and read this article below

[Thariq @trq212]

You can also see example HTML documents I've generated here: https://t.co/QuTMbHtdlc

[Andrej Karpathy @karpathy]

Fireside chat at Sequoia Ascent 2026 from a ~week ago. Some highlights:

The first theme I tried to push on is that LLMs are about a lot more than just speeding up what existed before (e.g. coding). Three examples of new horizons:

1. menugen: an app that can be fully engulfed by LLMs, with no classical code needed: input an image, output an image and an LLM can natively do the thing.
2. install .md skills instead of install .sh scripts. Why create a complex Software 1.0 bash script for e.g. installing a piece of software if you can write the installation out in words and say "just show this to your LLM". The LLM is an advanced interpreter of English and can intelligently target installation to your setup, debug everything inline, etc.
3. LLM knowledge bases as an example of something that was *impossible* with classical code because it's computation over unstructured data (knowledge) from arbitrary sources and in arbitrary formats, including simply text articles etc.

I pushed on these because in every new paradigm change, the obvious things are always in the realm of speeding up or somehow improving what existed, but here we have examples of functionality that either suddenly perhaps shouldn't even exist (1,2), or was fundamentally not possible before (3).

The second (ongoing) theme is trying to explain the pattern of jaggedness in LLMs. How it can be true that a single artifact will simultaneously 1) coherently refactor a 100,000-line code base *and* 2) tell you to walk to the car wash to wash your car. I previously wrote about the source of this as having to do with verifiability of a domain, here I expand on this as having to also do with economics because revenue/TAM dictates what the frontier labs choose to package into training data distributions during RL. You're either in the data distribution (on the rails of the RL circuits) and flying or you're off-roading in the jungle with a machete, in relative terms. Still not 100% satisfied with this, but it's an ongoing struggle to build an accurate model of LLM capabilities if you wish to practically take advantage of their power while avoiding their pitfalls, which brings me to...

Last theme is the agent-native economy. The decomposition of products and services into sensors, actuators and logic (split up across all of 1.0/2.0/3.0 computing paradigms), how we can make information maximally legible to LLMs, some words on the quickly emerging agentic engineering and its skill set, related hiring practices, etc., possibly even hints/dreams of fully neural computing handling the vast majority of computation with some help from (classical) CPU coprocessors.

[Elon Musk @elonmusk]

Grok 4.3

[CyrilXBT @cyrilXBT]

ANTHROPIC JUST EXPOSED HOW BADLY MOST PEOPLE ARE PROMPTING CLAUDE.

Their applied AI team dropped a 24 minute workshop.

Free.

From the people who wrote the model.

Not a course creator.

Not someone who figured it out by accident.

THE TEAM THAT BUILT THE THING.

Here is what makes this uncomfortable to watch.

There are 6 elements to a properly structured Claude prompt.

Most people are using 1.

Maybe 2 if they are being generous with themselves.

That gap is the difference between Claude giving you something useful and Claude giving you something you could have Googled.

The people who watch this workshop tonight will prompt differently tomorrow morning.

The people who skip it will keep wondering why their outputs feel slightly off no matter how much they tweak the wording.

24 minutes.

Free.

From the only people on earth who know from the inside exactly how Claude thinks.

I watched it twice.

Then I built a Claude Skill that applies all 6 elements automatically so you never have to think about prompt structure again.

Every prompt you run goes through the framework without you doing anything manually.

Full guide and the skill setup is below.

Bookmark this.

Come back to it this weekend.

This is the thing that compounds.

Follow @cyrilXBT for the exact Claude skills, prompt architecture, and systems I use to get outputs that most people do not believe came from one person.

[Khairallah AL-Awady @eng_khairallah1]

🚨 Anthropic's own team just showed how to actually use Claude Code properly.

30 minutes. free. the person who created Claude Code.

watch the workshop. bookmark it.

worth more than every $500 course you almost bought.

you've been using Claude without knowing 40 of its commands.

Then read the guide below.

[Andrej Karpathy @karpathy]

Wow, this tweet went very viral!

I wanted share a possibly slightly improved version of the tweet in an "idea file". The idea of the idea file is that in this era of LLM agents, there is less of a point/need of sharing the specific code/app, you just share the idea, then the other person's agent customizes & builds it for your specific needs.

So here's the idea in a gist format: https://t.co/NlAfEJjtJV

You can give this to your agent and it can build you your own LLM wiki and guide you on how to use it etc. It's intentionally kept a little bit abstract/vague because there are so many directions to take this in. And ofc, people can adjust the idea or contribute their own in the Discussion which is cool.

[Andrej Karpathy @karpathy]

LLM Knowledge Bases

Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So:

Data ingest:
I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them.

IDE:
I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides).

Q&A:
Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale.

Output:
Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base.

Linting:
I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into.

Extra tools:
I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries.

Further explorations:
As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows.

TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

[Elon Musk @elonmusk]

😂

[Akshay 🚀 @akshay_pachaar]

https://t.co/SSSIK3BX4z

[tengyuan @jutou]

@TechWith_Nova Banana

[NOVA @TechWith_Nova]

Claude Code + Nano Banana 2 is f*cking cracked 🤯

I built a skill inside Claude Code that writes JSON image prompts for Nano Banana 2, and the outputs look like they came from a professional photo shoot.

One plain-text prompt. Claude rewrites it as structured JSON with lighting, camera, composition, style, and negative prompts.

Then fires it off to Nano Banana 2.

All inside Claude Code.

Perfect for DTC brands and agencies who need high-volume ad creative without booking a shoot.

If you're using Nano Banana 2 for product shots and lifestyle images but every generation feels like pulling a slot machine lever — random lighting, inconsistent style, plastic skin, misspelled labels ...

This skill fixes the entire output:

→ You describe what you want in plain English
→ Claude rewrites it as a structured JSON prompt (lighting, camera angle, lens, depth of field, color grading — all of it)
→ Fires it to Nano Banana 2 via API
→ Saves the prompt + image in organized folders
→ You iterate on the style until it's dialed, then every output matches

No more slot machine prompting.
No more inconsistent brand imagery.
No more burning credits on unusable generations.

What you get:

- Photo-realistic product shots and lifestyle images on demand
- Full control over style, lighting, composition, and camera settings
- Saved JSON prompts you can reuse across every campaign
- A skill that gets smarter the more feedback you give it

Built 100% in Claude Code with a custom skill + Python scripts.

I put together a full playbook showing the exact skill, the JSON schema, and the workflow to set this up yourself.

Want the full playbook?

> Like this post
> Comment "BANANA"

And I'll send it over (must be following so I can DM)

[Miles Deutscher @milesdeutscher]

If you read just ONE article this weekend, make it this.

A 5-minute read could genuinely change your career trajectory.

While everyone else panics about AI job losses, this is the blueprint to ensure you thrive.

I wrote this for friends & family first - now it's public:

[Viktor Oddy @viktoroddy]

I recorded a 18-min tutorial on The New Google Stitch - Better than Figma (And It's Free)

[Sameer @samtwtss]

bro, it's so over for designers

google stitch is insane. 🤯

[tengyuan @jutou]

Support Z-Library https://t.co/FloC87UXqL

[tengyuan @jutou]

https://t.co/nvNBsJjdtY

[tengyuan @jutou]

https://t.co/nBgL6FoDQo

[tengyuan @jutou]

What is Supply Path Optimization?
https://t.co/Lf1N3M2f3X

[Peter Yang @petergyang]

Airbnb has a blueprint of their entire customer journey on their office walls.

They then map all product, policy, and service updates to it.

More companies should do this to avoid shipping the org chart.