Musaj.

Fullstack Developer, Building @AirbillsDigital

[Musaj. @Iam_Musaj]

Airbills aspires to become a Decentralised Hub for Research & Development on economic incentives, coordination games, and novel marketplaces.

Find us at the next round of Gitcoin Grants as a guild under MotionwaresDAO.

#gr15

@gitcoin
@ProtoResearch
@Molecule_dao
@KERNEL0x

[Gergely Orosz @GergelyOrosz]

@pommedeterre33 @fnthawar pick and choose!

"clearly ahead" is the one on modern eng orgs. Eg. how Uber uses AI for development: https://t.co/gvLGq90qz0

and the 3x "lessons from building" at The Pragmatic Summit, all cutting edge https://t.co/peoud8jFgs

[Stitch by Google @stitchbygoogle]

Let’s build a shared and open visual language for AI design. 🤝

Check out the open-source spec, try out the CLI, and let us know your thoughts on the new Components workflow!

🐙 GitHub Repo: https://t.co/qpqGcm7bU6

📰 Read more on The Keyword: https://t.co/LymOqg4Cw4

[Elon Musk @elonmusk]

You can access 𝕏 APi via @OpenClaw.

We’re trying to make it affordable without giving away the shop.

Hopefully, this can be useful & fun 💫

[Tom Blomfield @t_blom]

To get LLMs to work effectively, you need to extract and document all the implicit "domain knowledge" in your business.

Once you do that, AI agents are magical.

But it also means that OpenAI/Anthropic have full access to your company's domain knowledge 😬

[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.

[Ivan Zhao @ivanhzhao]

The loudest story about AI is a lonely one. One person with an army of chatbots. Other humans are friction.
That gets the future wrong. The best things aren’t built alone.
In a moment of change, we want to remind the world (and ourselves) what Notion stands for:
— Think Together

[Musaj. @Iam_Musaj]

@shankorashidi @abazmmanuel 🌚

[primordial intelligence @primal_brainer]

The older you get, the more you realize luck is mostly exposure.

If you sit in the same place, have the same routine, talking to the same people, nothing new really happens.

You have to tackle the world to win.

Travel more. Talk to people. Try a breakfast spot. Post on social media. Start a side hustle or a hobby.

The world rewards motion. You don't find opportunity sitting still.

[DAN KOE @thedankoe]

https://t.co/xDEP4ad1tz

[Andrej Karpathy @karpathy]

New supply chain attack this time for npm axios, the most popular HTTP client library with 300M weekly downloads.

Scanning my system I found a use imported from googleworkspace/cli from a few days ago when I was experimenting with gmail/gcal cli. The installed version (luckily) resolved to an unaffected 1.13.5, but the project dependency is not pinned, meaning that if I did this earlier today the code would have resolved to latest and I'd be pwned.

It's possible to personally defend against these to some extent with local settings e.g. release-age constraints, or containers or etc, but I think ultimately the defaults of package management projects (pip, npm etc) have to change so that a single infection (usually luckily fairly temporary in nature due to security scanning) does not spread through users at random and at scale via unpinned dependencies.

More comprehensive article:
https://t.co/EJAZbqAPIQ

[Eric Jorgenson 📚 ☀️ @EricJorgenson]

🚨📕 THE BOOK OF ELON IS NOW LIVE!!! 🎉🚀

This is the book we WISHED @elonmusk would write…

“All of Elon's most useful ideas, in his own words.”

Learn directly from the world’s greatest entrepreneur, like you’re sitting across from him at dinner.

It took FIVE YEARS to make this for you.

Because it's built from hundreds and hundreds of Elon's public appearances.

I went through 3,000,000+ words to collect the most useful and timeless ideas.

The final book is ~50,000 words.

Every word is USEFUL.

(This is what I do. My first book, The Almanack of Naval Ravikant, is one of the top 100 most highlighted books of all time on Kindle.)

Then, I spent $50,000+ on editing and design so it looks and feels beautiful.

Then…

> Foreword by @naval.
> Visuals by @jackbutcher.
> Blurb from @mrbeast.
> Published by @scribemediaco.
> And yes, approval on this idea from Elon himself, thanks to @samteller.

I went Maximum Effort to make this an all-timer.

We got 10/10 on reviews from early readers, then worked on it for ANOTHER YEAR.

Why so much effort?

My mission is to create One Million Musks.

For a generation to lift our gaze and build, so our grandchildren live in a world beyond our wildest dreams.

I’m an independent author.

I don’t get an advance.

I risk my own time and money to make these books.

Then we give away millions of them. Digital versions are free.

I believe this book can benefit every human, and if you can’t pay five bucks for it, I want to personally gift it to you.

Because I know it is useful.

Useful how?

You may be seeking purpose, a mission worthy of your life’s effort.

You may have a clear purpose and seek the tools for success.

You will find both in this book.

Get the benefits of Elon’s entire life of hard-won lessons in a five-hour, easy read.

(I checked, it’s a 5th-grade reading level.)

You’ll feel personally mentored by the greatest entrepreneur in history.

Click below to buy it now on Amazon, Audible, or directly from me.

Amazon: https://t.co/QfAT4e6unY
Audible: https://t.co/BSCkUBaPw8
Me: https://t.co/A7ckDxv7Dk

If you’re not sure it’s worth $4.99 yet, just start reading the free version.

PLEASE take 6 seconds to Like, Bookmark, and Repost.

Even better: send this to your friends, team, or Group Chats!

I guarantee this book will improve their lives.

Spread the word!

Every little thing helps.

Your support spreads good ideas around the world, helping people and making the future better for everyone.

Thank you!

Forward. Together.

[Andrej Karpathy @karpathy]

Software horror: litellm PyPI supply chain attack.

Simple `pip install litellm` was enough to exfiltrate SSH keys, AWS/GCP/Azure creds, Kubernetes configs, git credentials, env vars (all your API keys), shell history, crypto wallets, SSL private keys, CI/CD secrets, database passwords.

LiteLLM itself has 97 million downloads per month which is already terrible, but much worse, the contagion spreads to any project that depends on litellm. For example, if you did `pip install dspy` (which depended on litellm>=1.64.0), you'd also be pwnd. Same for any other large project that depended on litellm.

Afaict the poisoned version was up for only less than ~1 hour. The attack had a bug which led to its discovery - Callum McMahon was using an MCP plugin inside Cursor that pulled in litellm as a transitive dependency. When litellm 1.82.8 installed, their machine ran out of RAM and crashed. So if the attacker didn't vibe code this attack it could have been undetected for many days or weeks.

Supply chain attacks like this are basically the scariest thing imaginable in modern software. Every time you install any depedency you could be pulling in a poisoned package anywhere deep inside its entire depedency tree. This is especially risky with large projects that might have lots and lots of dependencies. The credentials that do get stolen in each attack can then be used to take over more accounts and compromise more packages.

Classical software engineering would have you believe that dependencies are good (we're building pyramids from bricks), but imo this has to be re-evaluated, and it's why I've been so growingly averse to them, preferring to use LLMs to "yoink" functionality when it's simple enough and possible.

[Name cannot be blank @hackSultan]

1. Tweet about what your users do alot.
2. About the product and what stage you’re in.
3. Once in a while, tweet about adding features that’ll stress out your engineering team.
4. About what’s going on with your investors.
5. About the kind of partners you’re looking to work with.
6. Start building the public image about the company you’re building and let people know that can ask your questions.
7. Get involved in the ecosystem you’re building in.

[Sora @soraofficialapp]

We’re saying goodbye to the Sora app. To everyone who created with Sora, shared it, and built community around it: thank you. What you made with Sora mattered, and we know this news is disappointing.

We’ll share more soon, including timelines for the app and API and details on preserving your work. – The Sora Team

[Claude @claudeai]

You can now enable Claude to use your computer to complete tasks.

It opens your apps, navigates your browser, fills in spreadsheets—anything you'd do sitting at your desk.

Research preview in Claude Cowork and Claude Code, macOS only.

[kagearchvs @kagearchvs]

the ability to execute every idea in your head brings a fulfillment i cannot explain

[Elvis @elvissun]

https://t.co/DotZ3V6XhJ

[Arpit Bhayani @arpit_bhayani]

No matter how many blogs, books, or papers you read, prototyping is still the fastest way to understand anything. Here's my workflow -

- keep a GitHub repo called "prototypes"
- every folder is one experiment
- something seems interesting, implement
- define exactly what you want to understand
- find the absolute minimum required
- code, run, iterate

The key thing here is finding the bare minimum, the absolute bare minimum, to code that would help you build that understanding. Initially, you will code more than required, but over time, you will start assuming data, adding sleeps, mocking, and making the right assumption.

Also, I get it, there is an urge to open-source the prototype or turn it into a project or startup. Don't. Remember, the goal is understanding. Once you get it, you are done; move on.

Yesterday, I built and implemented different types of joins (article published today) and benchmarked them to see how they actually performed. I already knew the theory, but the prototype gave me real, rough numbers.

By the way, I have 200+ repos on GitHub filled with different types of prototypes and hands-on experiments. Of course, I cannot recommend smashing your fingers on your keyboard enough.

[Musaj. @Iam_Musaj]

🎯

[Fireship @fireship_dev]

I've been a bit quiet here lately, but it's about time for an update: Fireship is officially merging with @uidotdev.

@tylermcginnis and the team make great courses and I've been a fan of Bytes since it launched. We actually talked about merging together back in 2021, but the timing wasn't right. Turns out it just took a few years.

For a little more context, ~18 months ago I sold a part of Fireship to a company called Electrify that invests in educational YouTube channels and helps build a team around them.

I did this for 2 reasons:

- Money
- Time

Most people don't realize that until recently, Fireship has always been a one-man show. I started with zero video production skill and zero expectations. I spent thousands of hours on a mission to make programming videos I’d actually want to watch. And somehow, I just kept making videos and it just kept working for the past 8 years…

I now have 3 kids with a 4th on the way. Running and scaling everything by myself isn’t sustainable anymore, and tbh, I never had the desire to manage people. I'm a hardcore introvert who just likes making YouTube videos.

So over the last year Electrify has helped me start building a team so I can focus on just the creative side. But the results have been mixed so far as we’ve tried adding to the team while still making content.

Despite the rumors, I have not been replaced by AI. If I had been, uploads would be way more consistent. The reality is that making this weird style of niche technical content is hard. AI can't do it, and it turns out most humans can’t do it very well either.

So we're changing things in 2026, and merging with the uidotdev team is the first step.

Going forward, we’ll be making more old school Fireship programming videos, more 100 Seconds videos, and more Code Reports on a wider range of topics, along with some crazy new ideas I think you'll like.

We’re also hiring more full time developers who are good at writing about technical topics. Editing and shitposting skills are also welcome. DM Tyler if you’re interested.

Yes, this all takes money, which is why we have more sponsors now. That’s the tradeoff for being able to build a team.

And no, I’m not stepping back. I'll still be voicing and working on every video, and I still have creative control to decide what we make. But the hope is that expanding the team will let us make more stuff and go deeper on everything without me burning out.

Also, existing Fireship Pro customers now get access to all uidotdev courses and vice versa. Check out the new fireship .dev site if you’re curious.

And finally, a huge thank you to everyone who has supported the channel over the years. I genuinely appreciate it more than I can put into words.