Ph.D. in Physics ⚛️ | AI & ML Expert. Leading innovations. InCredibles & Innowator Śląska 2024 winner. Bridging deep science & business impact.
Poland, San Jose USA
- Tweets 154
- Following 177
- Followers 88
- Likes 195
I’m building something awesome. 👀
The inspiration came while on holiday with my kids—but this is the next chapter of my AI-native device project, Vault.
I’ve tested the stack with current SOTA open-weight LLMs, and it works. The ultra-compact enclosure for my custom Jetson Orin Nano Super build is now printing. Fully parametric, too—no Astra operating Blender involved. 😉
Probably my most ambitious after-hours Physical AI project yet. Still need to add a physical kill switch… just in case! 😅
I’m fascinated by how far I can push this.
Where it started:
callstack.com/blog/building-…
#PhysicalAI #OpenSource #Hardware
2
5
132
This is your tool for testing mobile apps with agents, any platform, simulators and devices included ⬇️
Any agent can test iPhone Duo apps for you now. No need to do it by hand.
Folded, half-open, open, all without stealing your focus or other agents sessions.
Just give them @agent_device
2
1
15
2,974
How does quantization affect APEX performance? 🔬
Artur Morys-Magiera @artus9033 and I shared our research at @AgentConf . We hope our findings help the community make more informed decisions about quantization.
🎥 piped.video/XLoY8PpYXNU
5
17
4,020
344 runs. 25 days. Quantization is humbling 😅
Apex-31B BF16: 57% → 86.5% MMLU after fixing the chat template + eval harness. Same weights. Same 400 questions. 🤯
Then the payoff: IQ4_XS gave us 3.4× decode throughput and a 73% smaller file vs BF16 on DGX Spark. 🚀
@artus9033 and I talked about all of this at @AgentConf , organized by @callstackio.
For me, it’s already one of the most exciting AI conferences in Poland! 🇵🇱🔥
1
4
12
625
I’m still so surprised that, besides being a speaker at @code_europe, I WON a @PlayStation 5! 🎮🔥
Huge thanks to @softswiss for the generous prize in their lottery!
The funniest part? When I first got the “Congratulations, you won!” email, I was convinced it was phishing. 😅
Turns out it was very real. What a great surprise after Code Europe!
3
11
384
Last preparations yestarday. Today @AgentConf begins! I will be at APEX and Agent Device booth at 12. We can talk about our latest achievements in AI and agentic workflows. Meet me there!
2
12
311
Heading to Warsaw today in a great mood ☀️ Tomorrow at @code_europe Code Europe, I’m speaking about “No Framework, No Server: Making AI Agents Collaborate in One Terminal.” A terminal, local models, and agents working together. If you’re coming, say hi!
@AiswaryaVenkit1 @lemanskills @amireslamdoust
@tripu @aurelievache @BassemDy @menzel42
@cristinaponcela @dvdstelt @Everlier
@jarek_gajewski @joshleecreates @k_cieslak
@mourjo_sen @niallm @pkivima @sliwaptr
@pladynski @Raj_tldr @blackgirlbytes
@ro14nd @edorado93 @soganmageshwar
@szymonchudy @TejasKumar_ @girba
Made with AI
1
2
11
346
Releasing free unlimited React Native cloud Simulators
An OSS CLI that hosts simulators on Github and lets you or your agent run @agent_device to test things out. Available for private repos too.
Try out the alpha today at reactnativefeel.com/sim or install right away w/ `npm i -g native-sim`
24
33
412
61,789
GPT-6 Astra vs Claude Fable 5.1 on RN Evals.
134 tasks. 10 runs each. 2,680 fully judged evaluations.
Published overall scores:
Astra: 86.4%
Fable: 89.0%
Categories: Navigation, Animation, Async State, Lists & React Native APIs.
A narrow lead for Fable. One run doesn’t tell the whole story.
rn-evals.vercel.app/
1
1
7
161
Agents Commander 0.1.5 is out.
Now MIT-licensed: up to 100 agent panels, OpenCode, local agent-to-agent messaging, and opt-in capture with reviewed training-data export.
Try the offline demo:
npx agents-commander@0.1.5 --demo
npmjs.com/package/agents-com…
Made with AI
2
8
550
