Founding Avionics Engineer @ Utopia Labs
San Francisco, CA
Joined August 2020
- Tweets120
- Following236
- Followers96
- Likes448
Stefanos Fragkoulis retweeted
The second flying car of the week.
This one is also in early stages: @UtopiaLabs which is in the @fdotinc San Francisco startup factory.
I share these because I just love people who are building things and companies.
And because the FAA has started allowing autonomous flying vehicles so entrepreneurs see a new market coming.
What you building?
Stefanos Fragkoulis retweeted
Every big tech company is building a personal agent right now.
But most consumers don’t have personal problems with 10x upside.
Within 6-12 months, every one of these platforms will pivot to work.
Coding agents took off because the value per task is high enough to justify the token spend multiple.
Work has the margin, the pain and the economic leverage. That’s where agents actually scale.
Stefanos Fragkoulis retweeted
We brought in a founder building synthetic neurons.
His goal is to sell the first artificial brain, to a robot.
He gave a brief tutorial on how to make one:
22:47 Building a synthetic neuron from scratch
24:15 Making a robot move using only synthetic neurons
28:33 Dendrites, cell bodies + axons with hardware
32:42 Turning sensor data → spikes → movement
34:11 The roadmap to a full synthetic brain
38:00 A breakthrough in synthetic memory
By the end, we weren’t really talking about robots anymore...
Stefanos Fragkoulis retweeted
let’s pick some strawberries with our arms @ihwylie @IsaiahBallah @Ryan_Resolution
If you need arms for your projects hit us up dms open.
Stefanos Fragkoulis retweeted
The Chinese already killed Jev. This one went fast
介绍比Jev快50倍，在你设备上跑的laya-mlx!
只在你的设备上占用最高1G内存
Laya是一个开源的类似于Jev的,基于文本输出概率的分类系统
我将其移植到MLX,并且做了一些性能优化!
视频中就是这个模型在我的本地M3Max上玩贪吃蛇
这个模型能够以每秒决策60次的速度玩贪吃蛇！
github.com/mizorewww/laya-ml…
Readers added context they thought people might want to know
Laya was developed by an Indian Researcher - Nandakishor M
Source :
laya.convaiinnovations.com
arxiv.org/abs/2503.23303
Stefanos Fragkoulis retweeted
I can't confirm any of this, but I'm at least hearing the rumors.
1) GPT-6-Sol will be another significant leap forward; its price-performance ratio compared to Astra will be surprisingly good. Most likely, it will be released on Tuesday.
2) Opus 5.5, in response to OpenAI's move, is also expected this week, today or tomorrow. A significant leap forward; OpenAI and Anthropic are currently neck and neck. Opus is a response to GPT-6-Sol.
What's new to me is that "Bel" is internally referred to as AGI. Considering the graph, which still shows how quickly OpenAI's models continue to improve, this isn't surprising (graph below).
Rumors I’ve been hearing, not here on X.
First, let’s start with OpenAI and I’ll go towards Anthropic.
GPT-6 Sol is coming Tuesday. It’s both cheaper and more intelligent than 6 Astra, think of it like a 6.2 jump. The internal model “significantly more capable than Astra,” named Bel internally, helped with this release. Bel is considered “AGI” within OpenAI. They are very impressed with this model.
OpenAI is growing very confident that their internal lead is so big that no other lab can catch up. Unbelievably confident.
Anthropic is currently not in, let’s say, a “code red,” but is aware of OpenAI’s lead and doing everything in their power to catch up. Their new model Opus 5.5 is coming probably Monday rather than Tuesday due to OpenAI releasing on Tuesday.
Stefanos Fragkoulis retweeted
I built a voice controlled computer-use for my mac using @typesafeai's Jev and it's INSANE how fast it is!
I can dictate "open the notes app and create..." and the app opens before I even finish my sentence.
Stefanos Fragkoulis retweeted
Def working on a Jev video (coming soon)
But…..I wonder…..what would it take to make a local model do what Jev does?
Yes, we can make any model simply output multiple choice answers….I put Jev head to head vs a local model and Jev killed it on speed. It’s good.
Stats ——> Jev 96% at 118 ms vs Qwen3.6-35B on my Mac Studio 93.5% at 359 ms (sorting 200 text messages into spam or not spam). Those 200 Jev calls cost me about a quarter of a cent.
It’s closed weights and uses a new training algorithm called RLCD. Who is going to crack this for local first?
Stefanos Fragkoulis retweeted
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution