I built and sold Frame.io for $1.3 billion to Adobe, and along the way, I received plenty of advice that turned out to be entirely off the mark. Here's a look at some of the most glaring misdirections:

1. 'As CEO, focus on executive business tasks.' This couldn't…

Snowflake just launched the largest open-source model yet.

A 482B parameter MoE.

17B active parameters and 128 experts, trained on 3.5T tokens.

This model is more open than others — even the data recipe is open-source!

#AdobeFirefly now supports Generative Expand, which allows you to easily resize or change the aspect ratio of your images!

# explaining llm.c in layman terms

Training Large Language Models (LLMs), like ChatGPT, involves a large amount of code and complexity.

For example, a typical LLM training project might use the PyTorch deep learning library. PyTorch is quite complex because it implements a very…

Have you ever wanted to train LLMs in pure C without 245MB of PyTorch and 107MB of cPython? No? Well now you can! With llm.c:
github.com/karpathy/llm.c

To start, implements GPT-2 training on CPU/fp32 in only ~1,000 lines of clean code. It compiles and runs instantly, and exactly…

Composing Software by Eric Elliott is on sale on Leanpub! Its suggested price is $34.00; get it for $25.50 with this coupon: leanpub.com/sh/RAiaYwax Eric Elliott #ComputerProgramming #FunctionalProgramming #Javascript #Software #Software Engineering #Software Architecture

POV: It's the day before release. There's no such thing as a 'small change' at this stage. Even a 2-line code adjustment can endanger the release with:

* Further design revisions
* Downstream code changes
* Code review
* Fix automated tests
* Quality and bug fixes

When teams estimate work for a product, there's a HUGE GIGANTIC ERROR that they almost always make.

What's that error? They only count the cost of building a feature. 'This feature will take 4 engineer weeks to build.'

That's just the beginning of the cost.

🧵

🤣 and it still didn't work half the time.

The future of blockchain is a swarm of one million microinvestors writing you a $5 check into the seed round and giving you 12 seconds of post-investment management.

There's a new king of open-source.

The new LLM from Databricks just beat Mixtral!

- 132B total params (16 experts), 36B active (4 experts)
- Trained on *12T* tokens
- 32K context length

From my initial tests, I'm really impressed.

Composing Software by Eric Elliott is on sale on Leanpub! Its suggested price is $34.00; get it for $25.50 with this coupon: leanpub.com/sh/vbvwMyWy Eric Elliott #ComputerProgramming #FunctionalProgramming #Javascript #Software #Software Engineering #Software Architecture

Mocking is a Code Smell by Eric Elliott link.medium.com/PWSYOJ0LhIb

We keep hearing about the red flags (🚩) in software developers.

How about the green flags?

🟩Willing to be mentored.
🟩Willing to mentor others.
🟩Adapts to change well.
🟩Communicates early and often.
🟩Gives honest feedback.
🟩Identifies risks.
🟩Protects the team.

There appears to be a mismatch between publishing criteria in AI conferences and 'what actually works'. It is easy to publish new mathematical constructs (e.g. new models, new layers, new modules, new losses), but as Apple's MM1 paper concludes:

1. Encoder Lesson: Image…

Foundation Agent: a roadmap to build generally capable embodied AI that acts skillfully across many worlds, virtual or real.

Project GR00T, the Humanoid robot foundation model, is a cornerstone for Foundation Agent. It's the North Star, the next grand challenge in our quest for…

If you want to provide API services to React components, HOCs are better than hooks because you can call the unwrapped component directly without the HOC wrapper & test without mocking. You can also more easily reuse the enhanced component just by wrapping with a different HOC.

Inheritance vs. meritocracy, or why you shouldn't let rich kids invest your 💰

Mutual fund managers from poor families outperform those from wealth. This is because it is easier to be hired and promoted as a mutual fund manager if you come from wealth, even if you are bad at it.

AI folks drastically underestimate what you can do with a few hundred *incredible* examples.

DeepMind SIMA: an agent that plays 7 games and 4 3D simulations by reading pixels and generating keyboard/mouse control. This was the original promise of OpenAI Universe in 2016, way ahead of its time. After 8 years, it's done right with the modern AI stack.

Things I like:
-…