On my path to self-employment by building small bets. Author: O’Reilly NLP book, AI courses with Uplimit, Product @podsumo, AI consulting @nuefunnel
Germany
Joined March 2009
- Tweets3.1K
- Following1.4K
- Followers548
- Likes4.3K
Sid retweeted
Yoshua Bengio and collaborators just released a goldmine 182-page survey on mechanistic interpretability!
"Understanding AI from the Inside"
This paper will take you from neurons & circuits to how interpretability can help with AI safety
It covers SAEs, circuit discovery, steering, monitoring, how mechanisms emerge during training, and much more!
Read this excellent paper led by @FazlBarez and @TonyWu1105: alphaxiv.org/abs/2611.08438
Sid retweeted
Introducing, 100% free charts kit for you guys.
Spent whole damn month working on this 62 chart components. Available in Figma, code, or connect your MCP.
Download here: mebius.design/freebies/chart…
Sid retweeted
There is an absolutely WILD story in this @GergelyOrosz podcast episode
1. An unnamed CTO of a cash-flow-negative company was debating doing layoffs. AI
2. He says "I don't want to hurt morale like that, I won't do it"
3. A few months later, he says to @GergelyOrosz that he's cash flow positive. WTF. How?
4. The company had been licensing a component from a licensing company that was critical for their business.
5. One day, a dev thinks "what if we got an agent to build our own?"
6. The agent does so well that they produce a BETTER, FASTER version of the component they were licensing.
7. They debate open-sourcing it, and they ask the licensing company about it
8. The licensing company panics, and says "we will pay you NOT to open-source this"
9. They also say "we'll also pay you extra so we can re-license it"
10. So a big outlay became a recurring windfall, thanks to the ingenuity of a dev asking the right question.
Individual contributors have never been more powerful. Think bigger.
youtu.be/mQu0zN4IvSI?si=P_v9…
Sid retweeted
There has been a bunch of really interesting experiments with activating individual neurons in open AI models, locking them into particular personas.
Here historian @ResObscura tries the same trick with a model trained only on data before 1930, and the results are delightful.
Inspired by @Sauers_ and @jfonsecarivera and other fun recent work with personas in AI models, I built a partial neuron atlas for Talkie-1930, the historical language model created by @status_effects and @AlecRad. Here's the link: talkieneurons.vercel.app
Sid retweeted
装了几个热门的 DeepSeek Harness 插件，感觉这生态现在他妈的蓬勃发展
主要是 DeepSeek 太便宜，能力又中不溜的很强。
现在很多人就只用 DeepSeek Harness 了，尤其是爱折腾的年轻人，感觉跟以前刷机、搞搞 ROM 的是一类人
Sid retweeted
We're not a PE fund, but we do have a portfolio of companies. A few things I've seen:
1/ Execs often start with "oh this is a fairly well scoped task that takes a few hours a week. Maybe you can start with that." This is the WRONG way to think about AI. And you basically have to reframe the convo every time.
A better way is to start with their vision: "In an ideal world, how would this company operate?" And then work backwards from that.
The solution often requires a LOT of data wrangling (consolidating, cleaning, getting unwritten rules out of people's heads) and a little bit of AI.
If you take the exec's first thought at face value, you end up saving some guy five hours per month. But you don't fundamentally improve the business.
2/ So many people are still writing code by hand! There are huge gains to be had in simply showing people how to use Claude Code or Codex.
That said, in many legacy businesses, the codebases are not ready for agentic coding. So you have to do a bunch of unsexy work before you can let your agents run crazy.
3/ Another semi low lift win is finding out what wins are happening in a portfolio company and repeating that across other companies.
It rarely happens because the person who got the win is trying to get their next win (no incentive to codify and share their knowledge).
And even if you have a person responsible for hoovering up that knowledge and disseminating they might not have the authority to actually implement it.
What we're hearing about AI in PE on calls this quarter:
1. Funds are putting their own AI engineers inside portcos.
2. Portco CFOs see more pull requests than ever, but the codebase isn't getting better.
3. Dev vendors that won't use AI are being asked why they're still being paid.
4. Funds want AI work priced against EBITDA upside. They ask about revenue per FTE.
5. Every portco runs its own AI experiment. None of it makes it to the next company.
6. New board seats come with one job: choose the AI vendors.
The funds making progress are getting one workflow working at a time.
Sid retweeted
to all my followers who believe in this, sincerely - it's bullshit.
this is an era of unprecedented technology and with it comes unknown challenges, it is like a new world map in a game where only paid promotional beta testers have been before.
what is the value of discovering a counterfactual now when people have stopped reading everything?
how reliable is this tech where you chat with a bot and it hallucinates 3 times a day but you never know on what? what is the cost of discovering such bugs late for your projects?
what is the cost of discovering problems/bugs when entire world is wearing the same glasses and everyone has same blindspots?
"there is no technical moat" - yeah, for you.
ofc you would think that if you are someone who has always just wanted to get by their day and do nothing beyond what is forced on you, then yeah, sure.
but if you tell a guy like me "everyone is offloading to the same machine to think, using the same overabstracted superficially verified systems, where agentic PR count is a porn to be consumed by those who think they are doing something". if you tell a guy like me that, i would say there is now an opportunity to be hercules amongst men. there is an opportunity to really use this tech around you to imitate what fabrice bellard could have pulled off in your niche if he cared as much as you.
there is an insane opportunity to be unlimited, your moat is starting your morning where others have stopped caring about something.
Sid retweeted
Brutal truth from @Lethain
“If you're in [middle] management (…) you sort of have to decide, am I retiring now or am I re-skilling?”
He did just this: step down from leading a large team (250 ppl) to be hands-on with a small team (25 ppl), ship code & re-learn the new “rules”
Why a 3x CTO went back to a 25-person team. From Will Larson @lethain, Imprint CTO and 4x book author.
"It's a hard time in middle management, right? There's a lot more competition for these roles. What I found personally very rewarding is getting back into the details of writing software.
When I left Carta, my team was around 250. When I joined Imprint, the team I was leading was 25.
This a unique moment to actually re-skill. If you're in management, I think right now either you have to get back into the details, or the world is changing so radically that the profession you know will not be valued in three years. You sort of have to decide, am I retiring now or am I re-skilling?
I think there has to be some openness to taking a backward step or a lateral step. What I knew I wanted was getting back into the details and relearning how to do software in the current age. That's what I believe is going to be the unlock to the longevity of my career."
Sid retweeted
Hi, we're Threadlocked, a small collective of young researchers and hackers scattered across India and APAC.
We're sick of watching an entire generation mistake salary packages and credentials for the pursuit of computer science itself.
We're looking for curious hackers and researchers who still dare to take the road less traveled by treating computing as a craft.
We grew up in obscure corners of the Internet and learnt from strangers who encouraged our curiosity, shared their knowledge and even helped us access resources we couldn't afford. Somewhere along the way, we found our people. Now that we're in a position to do the same for others, we'd like to return the favor.
If any of these feel familiar we’d love to hear from you.
Sid retweeted
In practice most people experience their job as a queue of tasks. Tickets come in, tickets go out.
And the multiplier question ("Why does this kind of ticket keep coming in?") only occurs to you if you step out of the queue long enough to look at it. And most queues are designed, accidentally, to never let you do that.
Sid retweeted
kimi is dropping open weights for a 3T param model and Mckinsey is still in month 7 of the copilot pilot
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
What an interesting analogy with dirigibles - I’d heard of Hindenburg but the fact that airplane technology was not always obvious is new to me.
I've argued that the current era of AI is like the exciting era of dirigibles in 1920s -- flying by floating, without using the principles of aerodynamic lift. Conveniently, the timeline matches up exactly by 100 years, so maybe we can draw some parallels about mishaps too?
blog.dileeplearning.com/p/we…
88
Is this some kind of a prank or are they starting a company together 😂 @threepointone and @trq212 leaving their respective orgs around the same time. Learnt so much from what they share here! All the best.
47
Sid retweeted
I’ve left Anthropic after more than a year of working on Claude Code and seeing it scale larger than I ever imagined.
I have really loved working at Ant and believe deeply in the mission, and most of all, the idea of a race to the top.
The idea that if we work hard and responsibly, we can raise the bar and inspire others to great outcomes. And if we all push each other to be better, things can be radically good. I will carry the spirit of this with me forever.
My colleagues at Anthropic are the most talented, thoughtful, competent, and earnest people I have ever met. I am glad they are at the wheel and trust them deeply.
To the community on X, thank you for everything. You’ve made this the most fun (and sometimes stressful) job I’ve ever had.
For me, it is time for a bit of a break before diving into something new.
If you want more excellent Claude Code content, you should follow @addyosmani , @lydiahallie, @_catwu, @ClaudeDevs and of course @bcherny who started it all.
Sid retweeted
we just shipped the last @Blender AI plugin you’ll install
> 500 models in one tab. image, video, 3D, audio, textures, worlds
> for humans or agents
> open source
> live now
Sid retweeted
I think many of us working in tech tend to forget that the reason software engineering vs other engineering fields pays (a lot!) more, is a lot more flexible (WFH/remote), and has better perks:
Because this field changes much faster than other engineering fields. The premium is because fewer people are able/willing to pick up new technologies, methodologies so quickly.
But the field had ~10-20 years of stability, and so lots of folks forgot this is why a premium exists. And that AI is just another tech wave: software engineers who pick it up and master it will generally fare better; those not will struggle a lot more.
Sid retweeted
Maps in Rive! This will be released as a public library for Rive so you can add it to any of your projects! It renders standard Mapbox Vector Tiles from any source (including Mapbox's own APIs). It's fully controllable, stylizable, and can be placed anywhere (in a nested artboard, rendered to a texture, etc). Uses new features like AnimaScript, Threading API, and Fetch. Everything you see is dynamically loaded and processed on the fly.
The camera API is modelled on the Mapbox Maps SDK, but also gives you full access to custom shaders, rendering Rive artboards into the maps, etc. It switches between our 2D vector renderer and 3D (GPU Canvas) based renderer depending on the camera alignment. You can also force it via a mode to use one or the other.
Sid retweeted
What happens when you heavily distil a model with millions of decision based samples? I am super excited to release Drex Max, it's so far the strongest decision model which is top #3 in almost all the public benchmarks we evaluated. It can handle text, images, and video frames.
As promised:
⚡ Within 100 ms per decision, and under 1 s even at 262K tokens
💰 $0.06 per 1M input tokens
Open router support added.
