Building @Autohandai , former @Microsoft, @GitHub, @AWS.
Joined April 2007
- Tweets 11,052
- Following 1,963
- Followers 5,208
- Likes 2,285
427 Photos and videos
Pinned Tweet
Jun 10
WWDC26 is great, but watching the videos of How Apple decided to squeeze 20B model and I stumble on their Architecture, Looks very similar to what we shared with them. Left ours @autohandai Mneme architecture, right Apple AFM design
Yes, we have been built our accelerators for Small Dense models to run in some of our customers that has Apple servers.
1
425
Igor Costa retweeted
Jun 10
Concentration of power, capabilities and economic wealth is the biggest risk in AI. We need open science and open-source more than ever!
111
479
3,091
160,206
Jun 9
If you sell shovels and you hit bedrock, do you complain that the ground is too hard and stop selling shovels or do you say we need loops of loops ?
1
87
Jun 8
I have plenty of funny Oceania VC stories. But not all investor stories are bad.
One Australian GP, managing a <$1b fund, who didn't invest in us regularly sends bug reports and product feedback. His feedback has been worth more than many term sheets. The best investor is still your paying customer. Not all money is good money.
I'm not building a company for the next fund cycle. I'm building a company outlive me for centuries ahead.
If you can’t make your company interesting to a VC who is literally only in business to fund startups, how are you going to make it interesting to recruits and customers?
1
1
238
Jun 7
Post training is funny. You think you got the strength and endurance required to check and turns out you overfit the whole weights.
Tiny models to learn and evolve with you is what makes me wake up everyday
1
803
Jun 5
Thanks for having me Melbourne. This week at AI Engineer Melbourne 2026 I had the chance to meet some incredible engineers, founders, researchers, and builders.
One of many highlights was seeing how people are using @autohandai Code in the real world. One engineer showed me 30 agents running in parallel on our platform.
Those are the little sleepless nights that remind you why you build.
Distributed knowledge wins!
1
4
1,558
Jun 3
Welcome little fella. It’s now available on @autohandai Code
Jun 3
Meet Gemma 4 12B!
A unified, encoder-free multimodal model designed to bring high-performance intelligence directly to your laptop, and released under an Apache 2.0 license.
Bridging the gap between edge efficiency and advanced reasoning. Here is what’s new with Gemma 4 12B: 👇
1
472
Jun 3
Thank you for having me @aiDotEngineer Melbourne, @swyx maybe next time we have a chance to meet.
1
2
279
Jun 1
I must admit, I’m pretty happy to be surrounded by like-minded people.
Jun 1
We’re here Melbourne for @aiDotEngineer this week.
@igorcosta is presenting a keynote on the next generation of memory for coding agents.
Why do agents forget?
What does persistent memory actually look like?
Visit us at Booth 12 and chat with the @autohandai team.
1
1
272
May 27
AI Software factory age begins
May 27
Today we are announcing Autohand Code Enterprise.
We started with a CLI because real software work still happens in the terminal, usually one bad deploy away from character development.
Now we are bringing that same foundation to companies.
autohand.ai/updates/autohand…
2
810
Igor Costa retweeted
May 21
Performance：Qwen3.7-Max performs strongly across benchmarks in coding agents , and improves massively in general-purpose agents. Qwen3.7-Max also demonstrates exceptional strength on the hardest reasoning benchmarks, and stands out in general capabilities and multilingualism.
11
30
484
126,459
May 21
we just crossed 150k DaU. @autohandai Code comes with all batteries included and open and supports 25 LLMs for coding, BYOK for now.
I truly appreciate this platform users. Thank you!
1
1
3
369
May 20
Alfie is a fellow countryman from New Zealand and is in the Bay Area this week. Friends in the Bay Area. Try to show him around!
May 19
Introducing Libramen.
The agent economy is here. Headless services and buy-now buttons are trivial; transactions are already happening.
But real-world services don't work that way.
They require a scoping and qualification process before a transaction can take place.
Libramen solves this so agents can transact autonomously with IRL services. Now your agents can do anything from booking an event space with catering for your hackathon to hiring a contractor to clean your pool.
Link below.
1
4
290
May 20
I didn't know that everyone now is building a version of github.com/autohandai/comman…
We never got the credit for it. We started this as early or the same week as @charlieholtz . But I'm glad that society is ready for something more powerful in this space.
We're not actively maintaining Commander anymore.
1
67
May 13
Open source wins! Every single time.
May 13
Introducing /goals
Download now on 0.9.1 because @OpenAI and @claudeai copied from us --auto-mode.
Available for free at autohand.ai/code/
1
2
383
May 9
Frontier labs are betting AGI models will be so good you won't ever want to customize them. We think different. Building on a closed platform means renting your intelligence. The landlord sets the terms. They can give notice at any moment that your fine-tuning lease will not be renewed. As AI natives, we think you should own your AI. Your data, your domain expertise, your moat. Start training today, Autohand Evolve helps you retain your IP.
May 8
OpenAI has announced they will be winding down fine tuning. I got the email today. Existing active @OpenAI customers can keep running fine-tuning jobs until January 6, 2027, but after that no new training jobs can be created. Existing fine-tuned models will still run, but only until the underlying base model is eventually deprecated.
I get the argument that newer models follow instructions much better, and that prompts plus RAG cover more use cases than before. But not all of them.
1
243
May 8
We're getting ready to start offering our models via ONLY our Coding agent.
One of the main features: No rate limit, local or cloud.
Stay tunned.
1
6
267
May 7
We love our developers and thank you for being so patient with us 🫶
May 7
Today we’re introducing the Autohand Code Agent SDK.
Autohand Code started as a CLI, but teams quickly asked for a cleaner way to bring the same agent into their own apps, products, and workflows.
80