Alex Balfanz
5K posts

Alex Balfanz
@badccvoid
@atomic_semi, co-founder @badimo, programming Jailbreak on @Roblox, prev autopilot intern @Tesla, Duke University.
building Katılım Mart 2010
484 Takip Edilen294.3K Takipçiler

if you're into things like eventual consistency, computational geometry, combinatorial optimization, high throughput networking, circuit design, intelligent tools, UX, novel visualizations-- reach out :)
Atomic Semi@atomic_semi
We are building Atomic Studio to match the speed of our fab. It’s fast, browser-based, and collaborative. Layout, schematic, and simulation. Here’s a side by side of Studio (left) and Klayout (right) moving through the same file.
English

@Roblox @NosniyRBLX freaking sweet profile. Awesome game, awesome devs 👑
English

@tobi lots of great UIs have been made for profilers (Tracy, Firefox Profiler, Instruments, etc.) ... and now all we want is raw data. CLI for probing and querying perf. Bring back the ASCII!
English

@ContextLostRBX want to come work on computational geometry for chip fabrication @atomic_semi? Come tour our lab
English

nanochat now trains GPT-2 capability model in just 2 hours on a single 8XH100 node (down from ~3 hours 1 month ago). Getting a lot closer to ~interactive! A bunch of tuning and features (fp8) went in but the biggest difference was a switch of the dataset from FineWeb-edu to NVIDIA ClimbMix (nice work NVIDIA!). I had tried Olmo, FineWeb, DCLM which all led to regressions, ClimbMix worked really well out of the box (to the point that I am slightly suspicious about about goodharting, though reading the paper it seems ~ok).
In other news, after trying a few approaches for how to set things up, I now have AI Agents iterating on nanochat automatically, so I'll just leave this running for a while, go relax a bit and enjoy the feeling of post-agi :). Visualized here as an example: 110 changes made over the last ~12 hours, bringing the validation loss so far from 0.862415 down to 0.858039 for a d12 model, at no cost to wall clock time. The agent works on a feature branch, tries out ideas, merges them when they work and iterates. Amusingly, over the last ~2 weeks I almost feel like I've iterated more on the "meta-setup" where I optimize and tune the agent flows even more than the nanochat repo directly.

English

you can just dm me if you want a job. im serious. you can work with @henrl and @szeloof and @badccvoid
English

@JMCgaming24 @badimo @badccvoid @asimo3089 @badccvoid I think we found where the Roblox trivia Kahoot got its wrong release date from... 😅
English

@FireAtacck Love this video! congrats on making a great piece of art :)
English


Yoo, thanks to everyone who watched my recent video and supported the channel, it was much more attention than what we initially hoped for. 🔥
We’re getting pretty close for The Forge’s release. 24 hours public test gonna be announced soon in our discord server. Discord.gg/theforgers
#roblox #robloxdev #theforge
English

Lessons learned designing+building a (mostly) printed jet engine from scratch:
- SLM metal prints require substantial post processing to achieve anything close to traditionally machined ones
- Low-cost turbine balancing methods are very primitive
- 🧵…
Alex Boniske@AlexBoniske
Tentative next steps for early 2026 if we continue: remachine the backup impeller for less pressure leakage,re-test, then build a turbine-burst capable test rig and attempt a run on kerosene
English





















