fxtentacle

Location: Kiel, Germany (CET)

Remote: Yes. EU hours; US-East mornings work. On-site in northern Germany.

Willing to relocate: No

Technologies: C++, Python, PyTorch/TensorFlow, CUDA/SIMD; computer vision (retrieval at scale, inspection, SfM/SLAM); German speech recognition on-prem; real-time audio and kernel drivers; embedded/edge inference; RL environments (PPO, verifiable rewards); the boring enterprise plumbing around all of it

Résumé/CV: https://hajo.me/cv/

Email: hn@hajo.me

I'm the fractional ImageRights CTO, looking for a 2nd startup that needs hands-on tech, scale, or AI help.

I have 18 years of experience across 5 companies in taking products from idea to paying customers. I've built ImageRights' web search and image-retrieval database (C++, in service since 2015, finds copyright infringements worldwide), DeutscheKI (100% private German AI and local ASR), and before that the Spatial Audio Designer, used in Hollywood. And a German image distribution start-up.

I have experience with AI research, adaptation, and training. I've built AI RLVR training tasks in the past. Won the 2021 Bomberland mid-season with PPO. I'm also open to northern German on-prem CV/ASR work in case you cannot ship data to any cloud.

Side interests: I built a Pump It Up pad with custom self-developed and self-manufactured sensors. Currently doing CAD design and policy training (with PPO) for a robot.

Recently I've been debugging MSFS2024 inside Proton: https://github.com/ValveSoftware/Proton/issues/8255#issuecom...

if he “had been made aware that OpenAI has scraped and trained on information that was behind a paywall,” the company would have required OpenAI “to retrain its models.”

... yeah, sure.

I got 3x Bose QuietComfort 35 2 by buying 1. They did honor their warranty, but my impression was that they plan for people not going through the effort of actually requesting a replacement item, even though the warranty says they can.

The first one, "Henckels", openly says that their stuff is from China.

But the second link, "Zwilling" says "Approved". But boy was I disappointed that their "Zwilling Peak" branded ceramic hybrid frying pan for $100 says "Made in China".

"Why is it such a big deal?"

It's used by OpenAI.

I would like to run a robotics company. But raising for that is super challenging, because you have high execution risk combined with low margins on the final product. It's like the opposite of what western VCs want to see. That's why this type of company is usually founded in China.

And I disagree with you. I'm pretty sure parents worldwide would love a clean-up robot. You can avoid a lot of the risks by making it work only for empty rooms, which is OK because you need to clean up and let the vacuum robot run anyway (when the kids are away sleeping in their bedroom). There's been multiple attempts at building DYI robots to vacuum LEGO bricks off the floor. (Because stepping on them at night is such an immediate pain point.) Unless your product is soaked in hostile behavior and surveillance tech, parents will probably be happy to buy an in-home cleaning robot, no PR needed.

I think the main issue here is that parents are typically busy, which makes them bad startup founders, which is why they cannot "scratch their itch" and turn it into a business.

=> home cleaning robots do not exist because western VCs don't like it and parents aren't usually startup founders.

Also, didn't Roomba already leak non-consensual bathroom videos?

"VCs in the US are too big now — only social media (prev) or Al (now) have the scale they want."

I agree.

(I don't work at Anthropic, but I've designed RLVR tasks)

My impression is that especially for long-horizon tasks like science, the harness is much more important than people give it credit for. Claude Code + Fable 5 seems to have a tendency to "give up", get stuck in a dead end, or claim things to be impossible. But using the Fable 5 API together with a custom harness, it'll happily try 200+ variants and fail its way towards the goal.

If you give the AI a way to give up, eventually it will. If you remove that option from the harness, then thanks to the non-determinism inherent to LLMs, you get to explore pretty much all related solution attempts.

My experience with Claude is that it suffers badly from “not invented here” syndrome. So probably it rebuilt something like llama from scratch and then 2 hours suddenly seems reasonable (if you don’t question the approach). And that’s the thing, someone with no experience isn’t going to question it.

"using Codex for inference" says the article

<s> I'm glad they are completely not creepy for a company that plans to make money by selling your private data. </s>

"Capture the daily moments that define you. Voice the hidden details and tell the story only you can tell."

"We may sell, transfer or otherwise share [..] including personal information"

Nuclear Toothpaste

https://en.wikipedia.org/wiki/Doramad_Radioactive_Toothpaste

“many of Hollywood’s problems can be traced back to a single economic theory of everything: the rent is too damn high.”

Watching the video, 11:37 [1] made me feel very uneasy about the validity of those results. CMA-ES optimization times in seconds? That only works if you do no physics, no collision checks, no trajectory roll-outs. (Otherwise you're looking at 300+ iterations with 2000+ lanes each with 1000+ physics steps each, so millions to billions of simulation steps.)

And that really makes me wonder, if 30 tokens for a robot can be so expressive that they replace 600 mio simulation steps?

Or the alternative would be that they just assume that the given trajectory is equally optimal for each robot, which to me seems like a HUUUUGE assumption. But if that's the case, then the results of this technique would be highly misleading, as they would tell you the robot is excellent for 1 example movement, while you wouldn't know that a tiny change to your example movement might make the situation much better overall with a different robot.

[1] https://youtu.be/TTyjvPVFbNw?si=gU8zTfUyXVIMyvyp&t=697

I like the idea of adding a fake cpanel subdomain for 169.254.169.254 so that script kiddies will start port-scanning their own hosting provider, which will likely get them flagged/banned.

I read "midlife crisis" as "old guy trying to impress much younger girls" and then a cheap-ish car modded to be fast, furious, and obnoxiously loud seems like it might just work. Plus it's probably an amazing toy that can also be enjoyed solo.

And as long as there's enough attractive women who are impressed by loud cars, there will be guys with loud cars. I also dislike loud cars. But I'm at a loss as to how one would fix the root cause. Pull requests welcome ;)

And, yes, I share your impression. That car is about him trying to enforce attention by subjecting other people to his car.

Humans will typically learn after you have forced them to apologise for the same mistake for 20 times in a row.

AI won’t.

Fully agree. I tried to refactor parts of a large code base with Fable+ultracode and it just keeps accidentally merging distinct concepts and making up explanations/reasonings that the code base did not contain.

For example, the code base contains a physical controller. It’s closed loop in that it can react in realtime to changes. But it’s a slightly untypical implementation because this one can even look into the future through simulations. But Fable does not understand that. Instead, I need to remind it every 30 minutes that this is closed loop. It keeps wrongly claiming that the controller was open loop and then based upon that it will make up constraints that don’t actually exist.

Congratulations on the launch. The propeller demo looks good :) BTW, customising gears (which are then laser cut from sheet metal) is also a common request from untechnical people who need CAD help.

How do you differentiate from other open source AI CAD solutions? I’ve also had a good experience with Claude Code and CadQuery, which is a Python toolkit for generating CAD designs based on OpenCascade. I.e. what’s your moat?

And given that you’re YC funded .. what’s your plan for millions in MRR?

I’m asking because I would like to launch some useful open source tools myself (but for PIM, not CAD) and I just can’t figure out how to make open source sustainable, let alone profitable enough for hockey stick revenue growth. Any tips? Book recommendations?

The grab app shows un-skippable ads. And some taxis even have ads "based on your interests" playing on a TV in front of your seat. I found their "no marketing" ;) to be almost too much.

The thing that nobody talks about is how many EU citizens already are working on frontier models, just inside US companies. That’s why the prohibition was so crippling for Anthropic.

And, btw, the bang per buck that I got from OVH was better than EC2.

This is an interesting approach with integer programming and then using an explicit solver. It’s probably very slow, but you only have to run this once and it produces the mathematically perfect result.

In the past, I got good results with trying to reduce the variance in entropy in-between tokens, which you can implement very easily by starting with each single character as its own token and then doing a greedy merge of the most numerous outlier token pairs in a loop until you reach your desired token count. https://arxiv.org/abs/2206.12693

6

“ and prove that Age of Empires II is functionally- and Turing-complete.”

SEEKING WORK: Fractional CTO as Freelancer (Kiel, northern Germany, own company => 100% remote & easy to hire).

I built 3 products from idea to paying customers. So I can help your startup with architecture, implementation plans, and writing specs that you need to keep vibecoding sane. Depending on whether I get into YC (or most likely not, applied with just an idea), I might have free time coming up in the upcoming months.

https://hajo.me/cv/

Me, too. I love that they have tables with return and warranty percentages so you can see which products other people kept and which products only keep breaking.

The last 2 LAN parties we didn’t play anything that didn’t work extremely well on my AMD CPU + NVIDIA GPU Pop!OS Linux workstation.

And I didn’t expect: VR streaming works flawless, too. I just had to buy a WiFi 6 usb dongle.

"Maybe one day they will get around to unsuspending my account." is increasingly how support feels with all big cloud companies.

I got burned with an attitude like this: unexpectedly, people who had downloaded my open source tool for free started expecting support. Some of them sent pretty unfriendly emails.