I love to build products. CTO @factorialapp
Barcelona
Joined May 2008
- Tweets1K
- Following192
- Followers641
- Likes485
Ilya Zayats retweeted
Spanish unicorn @factorialapp built an agent across 25+ products for 17k+ customers.
The toughest part: configuring perms to access HR data. They had the agent call the same GraphQL API as the frontend, essentially inheriting user permissions.
Here's CTO @somebody32:
Ilya Zayats retweeted
You are an AI engineer now. That means you need to deeply understand the toolchain underneath you:
1. Chain of thought
2. Traces
3. Harness design
4. Benchmarking/evals
5. Sandboxing, safety/control
So many people are just trying to shortcut everything by throwing xhigh thinking at every problem and installing skillslop repos or agent memory projects that promise token savings or to 'make your agent smarter', whatever.
I see so little curiosity for how anything works, and so much laziness of people just wanting to eat at the trough of whatever OpenAI or Claude or X throws down the pipe at them.
If you say you're an engineer, act like one!
The more rumors there are about a model dropping next week, the dumber the current generation feels.
Not because it actually is, but because that’s how the hedonic treadmill works: you’re way more aware of the limitations, but you’ve already gotten used to the magic
I really like the capability split between 5.6 Luna and Sol, but I have a hard time understanding Terra's positioning.
In my experience, Luna is just a notch less capable on properly defined tasks, but dramatically cheaper. And for vague tasks you need to reach for Sol anyway
My TS AI Conf talk is live. I try to convince you that investing in agent tools pays off orders of magnitude more than tweaking prompts or constantly switching models
youtube.com/watch?v=dGPYsci6…
Every six months start from scratch: keep the evals, but rebuild everything else.
Models improve so fast that the cruft required by the previous generations becomes a liability
The current reality of 25 reasoning levels per model that force you to pick between correctness and speed is a clear signal that we are still in the Stone Age of AI
OpenAI has been nailing not just the models lately, but community building too
Updates for Codex and ChatGPT Work users. No nerfing, only good stuff!
- We have landed inference optimizations and are passing down savings to all the subscriptions for GPT-5.6 Sol. That should result in around 10% more usage on its own.
- We noticed that by changing the context size limit in the product to 372k for GPT-5.6 Sol, up from 272k for GPT-5.5, it resulted in more usage being charged than intended. We have reverted to 272k and will work to roll back out to 372k in the days to come. You should notice that usage drains significantly less after this change.
- To understand where the extra usage was coming from, we ran some experiments where reasoning efforts were changed (referred to as juice values under the hood) and have reverted this.
- There is slightly more usage of multi-agent than intended in high and xhigh reasoning effort, we are fixing this going forward. Also fixing a small other thing we noticed with auto-review where we can be more efficient.
And we continue to have the 5h limit temporarily not apply. Enjoy the rest of the weekend!
Ilya Zayats retweeted
Rewriting Bun in Rust
bun.com/blog/bun-in-rust
Goal in Codex is a blunt instrument that can wreck a codebase by overfitting to the objective.
But with proper back-pressure, waking up to a huge migration one-shotted overnight is deeply satisfying
The Catalan robot finally has a body!
Today I soldered the main parts and connected them to the “brain”.
The next step is to package everything and mount it on top of the octobot.
Got heavily nerd sniped by @badlogicgames. Never thought that building a robot together with a 5yo could be so much fun
Ilya Zayats retweeted
BREAKING: Spanish @factorialapp has JUST raised a WHOPPING $150m Series D at $2.5bn valuation! 🇪🇸
The round was led by @generalcatalyst (@pranavsinghvi), who have also committed an additional $540m through its Customer Value Fund.
This is AMAZING news for one of the top companies coming out of Spain.
I spoke to Founder and CEO Jordi and discussed:
0:00 - Introduction
0:21 - General Catalyst invests $150m in Factorial
1:10 - Why raise capital now?
2:28 - The relationship with General Catalyst
3:55 - How the Customer Value Fund works
5:24 - From HR software to AI-native workforce operations
9:00 - Leading AI transformation inside a company
11:18 - Why Europe is a great place to build
13:16 - Barcelona’s startup ecosystem
15:52 - Lessons from Silicon Valley
18:21 - Building a company for the long term
19:57 - What Factorial looks for when hiring
Congrats @jordiromero, @bernatfarrero, Pau Ramon Revilla, and the rest of the Factorial team!
Ilya Zayats retweeted
What we are losing with AI is syntax -- and good riddance. The less our brains are occupied by semicolons and braces the better. There are much more important things for us to consider and manage.
Seeing this more and more everyday. The difference is stark
Wow, I totally disagree with this statement. At the current state, AI actually amplifies the developer to developer difference. If you were a 10x developer, you had good ideas + architectural clarity, this is a brutal advantage when using AI. Steering is a fundamental part of today's AI development.