HNHacker News
TopNewBestAskShowJobs

fschuett

478 karma · joined March 19, 2022

submissionscomments
fschuett··on AI World Clocks
Reminds me of this: https://www.youtube.com/watch?v=OGbhJjXl9Rk
fschuett··on Ask HN: Who wants to be hired? (November 2025)
Location: Germany

Remote: Preferred

Willing to relocate: Anywhere, as long as the salary matches the local living costs, I'm not bound by family or similar

Technologies: I have about 6 YOE in Rust, so that's my daily driver and it's what I'm best at. Other than that, Python, general frontend tech (JS / HTML / CSS, but I don't use frameworks), SEO and graphic design

Resume/CV: https://drive.google.com/file/d/17pG-DUswO19foUCrctVR40jE5S-..., https://www.linkedin.com/in/fschuttmaps4print, https://github.com/fschutt

Email: See Resume.

I am mostly a self-taught Rust engineer who tried (unsuccessfully, so far) to build a cartography startup. But I still need a regular day job, even if everything goes well, but I have lots of experience in Rust by now. After school, I only had some training in government administration, but never wanted to work there. Learned C++ in my free type, then switched to Rust in 2017 to create my own GIS / graphics applications. That work resulted in the https://azul.rs GUI framework (still working on it).

I finished multiple smaller projects for the German government (had to get "some" job because I ran out of money building Azul), but sadly all my designs got rejected due to German bureaucracy. I am good at thinking through systems, ownership and understanding the entire product, as well as rendering, performance, Linux / open-source and software architecture.

Not interested in crypto or defense sector.

fschuett··on Free software scares normal people
Well, Tantacrul did answer that objection: it just shows you a popup dialog on first start: "which theme do you want" (colorful or colorless, light / dark) and "which experience do you want" (classic / new). So if you pick the "colorless, light, classic" option, it's going to look pretty much like the current Audacity, except that they moved from wxWidgets to Qt.
fschuett··on Free software scares normal people
> Free audio editing software that requires hours of learning to be useful for simple tasks.

To be fair, the Audacity UX designer made a massive video about the next UX redesign and how he tried to get rid of "modes" and the "Audacity says no" problem:

https://www.youtube.com/watch?v=QYM3TWf_G38

So this problem should get better in the future. Good UX (doesn't necessarily have to have a flashy UI, but just a good UX) in free software is often lacking or an afterthought.

fschuett··on Affinity Studio now free
If possible, please make a Linux version.

Just in case any Canva engineer is reading this.

fschuett··on Rust cross-platform GPUI components
Azul.rs is making lots of progress now and it'll have decent C / C++ / Python bindings (so it won't be Rust-only). But I guess for now GPUI is more mature.
fschuett··on Programming with Less Than Nothing
What inane, impure nonsense. You are still using "execution" and "output", these filthy, contingent effects of reality. Why would you use THREE entire combinators, when you could've just defined ιx = xSK. Pathetic. A true, pure program never needs to run. It just exists as a proof, and by its existence as such it simply reconfigures the consciousness of the user to align with its obvious, Manichaean-esque purity. "Running" the "program", please. What are we, barbarians?
fschuett··on Servo v0.0.1
Thanks, it's not really ready yet though, the screenshots were from 2019.
fschuett··on Servo v0.0.1
I know about Dioxus / Blitz, but it's a very, very different project. The only common part is that both Azul and Blitz use taffy for flexbox / grid, but both the technologies, architecture, funding and goals are extremely different:

Blitz:

- Custom renderer (Skia?) vs Azuls WebRender fork (to get rid of any C dependencies)

- Stylo (CSS parser) vs azul-css (to support compilation of CSS to const items)

- HarfRust (font shaping) - vs allsorts (I used allsorts also in printpdf, so it fits)

- Skrifa (font parsing) - vs allsorts again (simplifies things)

- Fontique (font selection) - vs rust-fontconfig (custom pure-Rust rewrite of fontconfig)

- Parley (line breaking) - vs Azuls text3 engine

- All as separate projects vs Azuls monorepo-style

Dioxus:

- RSX macros, data + function coupled together vs Azuls "C function callbacks + HTML dataset" model

- Binary hot-patching vs Azuls dynamic linking model

- Macros vs Azuls HTML/CSS to Rust/C compiler build tool (no macros)

- Funded by YC (not sure about upsell?) vs funded by donations (once it's stable enough) and my Maps4Print cartography startup (dogfooding)

These things matter, even for small decisions. For example, Azul uses a custom CSS parser because the CSSProperty is a C-compatible enum, so that later on you can compile your entire CSS to a const fn and use CSS strings without even doing any allocations. So even on that level, there's a technological-architectural difference between Azul and Stylo.

But the core point is more architecturally: Azuls architecture is built for de-coupling the user data from the function callbacks, because I see this as the Archilles heel that all GUI systems so far have failed at:

https://github.com/fschutt/azul/blob/master/doc/guide/02_App...

Dioxus however repeats this exact same pattern again, and even the Elm architecture doesn't really fix it. I didn't finish the document but basically there is a (1) "hierarchy of DOM elements" and a (2) "graph of UI data" and those two are not always the same - they can overlap, but the core assumption of many GUI toolkits is that (2) is a tree (it's a graph, really) and (2) is always in the same hierarchy as (1), which is why GUI programming is a pain, no matter what language / framework. Electron just makes the visual part easier, but then you still need React to deal with the pain of data model / view sync.

I can collaborate on the flex / grid solver ofc, but it's very hard to collaborate on anything else because the technologies used, the goals, the architecture, etc. are very different between Dioxus / Azul. Azul is more "monorepo-NIH integrated solution" (because I often got bug reports in 2019 that I couldn't fix because I didn't own the underlying crate, so I had to wait for the maintainers to do another release, etc. - I learned from that).

As a note, the layout engine is also now heavily vibe-coded (sorry not sorry), so I don't take credit - but feel free to take inspiration or copy code. Gemini says the solver3 code is a "textbook implementation", take that as you will. My idea was to build a "AI feedback loop" to semi-automatically put the HTML input, the debug messages (to see what code paths are hit), the source code and the final display list into a loop to let the AI auto-debug the layout engine. So that part of writing the HTML engine isn't really hard, assuming the plan works out. The hardest part is caching, scrolling, performance debugging, interactions between different systems, and especially supporting the C API. Layout is comparably simple.

fschuett··on Servo v0.0.1
You would end up simply with Electron 2.0. I tried de-entangling the Servo CSS / JS / Layout engine some years ago, to see if it would be more lightweight, it wasn't: https://github.com/fschutt/servo_gui_test (62 MB binary size, several hundred MB RAM usage IIRC)

I am currently working on getting https://azul.rs/reftest ready, which uses some of the underlying technologies as Servo (taffy-layout, webrender) but uses no JavaScript and also has a C / Python API. Azul is basically that, except it's not usable yet.

fschuett··on The World Trade Center under construction through photos, 1966-1979
Was the WTC 7 / Salomon Brothers Building part of the same construction?
fschuett··on Kurt Got Got
Hmm, since Chromium is working on adding browser-local AI features, I wonder if this one day could be a security check (for links opened from the outside of the browser). E.g. the browser detected that you clicked on a new-tab link, and the page looks like a commonly known site, then the AI detects that the URL isn't "x.com" and gives a heads-up warning. At least for the top 1000 most common sites, this could prevent a lot of phishing attacks.
fschuett··on Ask HN: What is nowadays (opensource) way of converting HTML to PDF?
Rendering to SVG, at least that's what I did on https://fschutt.github.io/printpdf/

I am currently writing a WASM-ready PDF toolkit that can handle both HTML to PDF and then rendering PDF pages to SVG. However, it's not yet production-ready.

The underlying HTML engine is currently a severe "work in progress", but it gives me the low-level access that I need: https://azul.rs/reftest

fschuett··on Replace PostgreSQL with Git for your next project
I did once write a system like that using libgit2 for the German Grundbuch (land registry):

https://github.com/projekt-dgb/dgb-server/blob/master/API.md...

In the current system, rights, owners and debts of any land parcel in Germany are simply recorded in PDF files, with an ID for each parcel. So, when adding a "record", the govt employees literally just open the PDF file in a PDF editor and draw lines in the PDF, then save it again. Some PDF files are simply scanned pages of typewriter text (often even pre-WW2), so the lines are just added on top. It's a state-of-the-art "digital" workflow for our wonderful, modern country.

Anyway, so I wrote an entire system to digitize it properly (using libtesseract + a custom verifying tool, one PDF file => one JSON file) and track changes to parcels using git. The system was also able to generate Änderungsmitteilungen (change notices) using "git diff" and automatically send them via E-Mail (or invoke a webhook to notify banks or other legal actors, that the parcel had changed - currently this is done using paper and letters).

It was a really cool system (desktop digitizing tool + web server for data management), but the German government didn't care, although some called it "quite impressive". So it's now just archived on GitHub. Their current problem with "digitization" was that every state uses different formats and extra fields for the and they are still (after 20+ years) debating about what the "standardized" database schema should be (I tried to solve that with an open, extensible JSON schema, but nah [insert guy flying out of window meme]). I'm a one-man show, not a multi-billion dollar company, so I didn't have the "power" to change much.

Instead, their "digital Grundbuch" (dabag) project is currently a "work in progress" for 20+ years: https://www.grundbuch.eu/nachrichten/ because 16 states cannot standardize on a unified DB scheme. So it's back to PDF files. Why change a working system, I guess. Germans, this is where your taxes are spent on - oh well, the project was still very cool.

fschuett··on Ask HN: Has anyone else been unemployed for over two years?
Technically 2.5 years here (I had a shitty job in the meantime for 4 months, but it wasn't a tech job). I have a decent GitHub portfolio though and I've made sure to never waste my time. So, I'm unemployed on paper, yet always "employed" building open-source software in reality. I have tried to finish building my cartography startup and my mini-browser engine. Once I have that, I can hopefully support myself with non-software money because the job market in Germany is a complete disaster, I still get interviews (about two per month), but they usually lead nowhere. Sometimes I get into the next round, just to be rejected for someone with more experience.

Sadly (?), I don't have any higher education and I'm too "self-employed" for corporate jobs (corporate jobs really, really don't like having someone build their own startup on the side). And on top I'm 26, not 36, so there's no way I'll have the experience required for someone truly "Senior". I get by on German social security, I get exactly 560€ / month and that's it (plus health insurance, 220€). If you wonder how someone can live on that low amount of money, it's because I accidentally inherited a paid-off house and don't need to pay rent (state covers any taxes, would be even more ludicrous if they didn't). So I have very, very few expenses, no liabilities and a few close friends.

I never wanted to be a drain on society (heavily socially punished in Germany), so I try to stay active and use my time for open-source projects. But since my net loss on society is relatively low anyway, I see it as morally justified to develop my "cartographic AI solution" while being a bum on paper. Let's just call it "government-subsidized startup seed funding". At least the thought of "finish your startup or you'll one day die of starvation" does do wonders for my motivation.

If people want to judge me for being on social security, I don't care anymore. I have my goals and I'm not running out of work, technically, despite being "unemployed". I care about building my skills and my startup and having "something for myself" so that I don't get financially torpedoed every few years (2008 crisis, 2015 crisis, Corona 2020, AI bubble 2025, ...). Once the job market gets better or I finish building my startup, I'll be better off. Until then I just have to deal with judging looks. How on earth someone is however supposed to build a stable family life from software engineering if the job market shits itself every few years is beyond me. I guess I lack the firm handshake and smile.

I did buy "Lingua Latina Per Se Illustrata", once I'm done with my programming projects, I'll focus on that. Some of my friends can speak fluent Latin, I gotta catch up. And learning math properly, working out, etc. But yeah, I'm lucky that I don't have many expenses. Stay active, don't waste time.

fschuett··on Stripe Launches L1 Blockchain: Tempo
The requirement "high motor" is a new standard in the software industry, where you are expected to arrive at the interview levitating at least five feet off the ground, propelled by your own internal combustion engine. Your resume should include your horsepower (minimum 250), fuel efficiency per Jira task and preferred brand of motor oil.
fschuett··on Go is still not good
Technically, the term "billion dollar mistake", coined in 1965, would now be a "10 billion dollar mistake" in 2025. Or, if the cost is measured in terms of housing, it would be a "21 billion dollar mistake".

:^/

fschuett··on Launch HN: Issen (YC F24) – Personal AI language tutor
Please add Latin
fschuett··on “Why is the Rust compiler so slow?”
For deploying Rust servers, I use Spin WASM functions[1], so no Docker / Kubernetes is necessary. Not affiliated with them, just saying. I just build the final WASM binary and then the rest is managed by the runtime.

Sadly, the compile time is just as bad, but I think in this case the allocator is the biggest culprit, since disabling optimization will degrade run-time performance. The Rust team should maybe look into shipping their own bundled allocator, "native" allocators are highly unpredictable.

[^1]: https://www.fermyon.com

fschuett··on Canon Law Ninja
Would be great to have English translations for the 1917 Code (esp. the canons regarding "what is a marriage", ... ). Right now I use Gemini for searching through the Latin text, it "works", but it would still be a decent value-add. Just in case the maintainer sees this. Also, semantic matching of 1983-1917 canons would be nice.
fschuett··on By default, Signal doesn't recall
> “Take a screenshot every few seconds” legitimately sounds like a suggestion from a low-parameter LLM that was given a prompt like “How do I add an arbitrary AI feature to my operating system as quickly as possible in order to make investors happy?”

No, actual AI is smarter than Microsoft managers, it seems:

Here are some ideas for adding an arbitrary AI feature to your operating system quickly to make investors happy:

- AI File Search: NLP for file/setting search (search files by NLP querying)

- Auto Window Layouts: AI-suggested window organization ("coding mode", "research mode" depending on detected usage patterns)

- Smart Notifications: automatic notification condensing to reduce clutter

- AI Clipboard: Keeping a categorized clipboard paste based on content

- Predictive App Launcher: Suggests apps based on daytime, usage, recently opened files

- AI Wallpaper/Theme: Smart visual suggestions, i.e. wallpaper based on current weather, mood, etc.

- Voice Quick Commands: AI-based voice OS control ("Open browser")

- AI System optimization: for example, content-based disk space cleanup

Any of the above are better than this nonsense.

fschuett··on OlmOCR: Open-source tool to extract plain text from PDFs
Very impressive, it's the only AI Vision toolkit so far that actually recognizes Latin and medieval scripts. I've been trying to somehow translate public-domain medieval books (including the artwork and original layout) to PDF, so they can be re-printed, i.e pages like this: https://i.imgur.com/YLuF9sa.png - I tried a Google Vision + o1 solution, which did work to some extent, but not on the first try. This even recognizes the "E" of the artwork initial (or fixes it because of the context), which many OCR or AI solutions fail at.

The only think I'd need now is a way to get the original font and artwork positions (would be a great addition to OlmOCR). Potentially I could work up a solution to create the font manually (as most medieval books are written in the same writing style), then find the shape of the glyphs in the original image once I have the text and then mask out the artwork with some OpenCV magic.

fschuett··on What if Eye...?
In order for this to work in real life, you'd have to prove a lot of other invariants:

- The mechanism to interpret the light data signal has to be in-step with the evolution of the eye. Getting light data without a brain evolving at the same time to interpret it is evolutionary recessive, i.e. a useless function. I.e. a real evolution would be more like "cat /dev/urandom > output.html", not a controlled ecosystem with a clear penalty-reward system.

- In nature, there is no 1:1 "reward / selection function" like in this simulation. In the computer, this "motivation factor" is externally given, so that the next generation is rewarded and selected out, in reality, there is no rule as to what is and isn't "better" or "fitter" or "more attractive to the other gender" (not like CS nerds would know). Sure, an organism can consume food, but beyond a certain point that wouldn't make the organism just "fat", not stronger. So there also need to be environmental mutations happening at the same time, that reinforce "more food = better evolved".

- There has to be a way for the animal to be so dominant, that the connection between light data and food can be genetically passed on and will not be associated with bad artifacts (see ChatGPT hallucinations for examples of "accidental bad artifacts in evolution" - and that "evolution" has millions of man-hours, money and R&D behind it).

- By the rule of "survival of the fittest", the next generation mutation has to be (in one single step) such a significant improvement over the last one that it won't be selected out again by recessive selection or dilution inside of the gene pool.

- The gene has to be active within 150 subsequent generations, without fail, cancer, recession and provide 150 times a dominant advantage, just to get a basic "eye" for 2D navigation with 10 light sensors. The minimum snail eye (pre-Cambrian) has 14.000 cells [1] (and a snail cannot see color).

- The real world is a 3D environment, which adds a monumental amount of complexity. Add to it the complexity of depth, color, shape, ...

- The mutation(s) have to happen either "at once" or be widespread (otherwise it's going to be like an Albino animal, i.e. some rare neutral mutation).

- All of this has to be done in an environment hostile to life in general (i.e. the edge of underwater vulcanoes, some primordial soup burning at several hundred degrees), all elements have to be at the right place, at the same time, etc. And be created out of nothing, of course.

While I do agree that it can be helpful for computer vision, computerized "evolution" is just adaptive statistical pattern matching, but it's absolutely nothing like real biology. It would be more realistic to just output "/dev/random > kernel-gen-xxx.iso" and then run it bare-metal, with no lab environment, no operating system, no programming language, no goal function, no selection / reward process, no debugging, etc.

Even Darwin had his problems with the eye. The reason I believe in God is not necessarily because I want to, but because evolution (not survival-of-the-fittest, but the "mutation creates information" aspect) requires far more faith and far more dogmas, which cannot be questioned for the sake of science. When I was in 8th grade biology, I took a stone from the schoolyard, put it on the teachers desk and said "alright, so this is a human if we wait 4 billion years". The teacher ignored me, but never told me I'm wrong.

[1] https://link.springer.com/article/10.1007/BF00606433

fschuett··on Cerebrum: Simulate and infer synaptic connectivity in large-scale brain networks
Simulating a brain would mean that reason, the ability to discern good from bad, is a statistical process. All scientific evidence so far shows that this is not the case, since AIs do not have the ability to "understand" what they're doing, their input data has to be classified first to be usable to the machine. Especially the problem of model collapse shows this, when an AI is trained on the output of another AI, trained on the output of another AI, it will eventually produce garbage, why? Because it doesn't "understand" what it's doing, it just matches patterns. The only way to correct it is with hundreds or even thousands of employees that give meaning to the data to guide the model.

Consciousness presumes the ability of making conscious decisions, especially the ability to have introspection and more importantly, free will (otherwise the decision would not be conscious, but robotic regurgitation), to reflect and to judge on the "goodness" or "badness" of decisions, the "morality". Since it is evident that humans do not always do the logical best thing (look around you how many people make garbage decisions), a machine can never function like a human can, it can never have opinions (that aren't pre-trained input), as it makes no distinction between good and bad without external input. A machine has no free will, which is a requirement for consciousness. At best, it can be a good faksimile. It can be useful, yes, but it cannot make conscious decisions.

The created cannot be bigger than the creator in terms of informational content, otherwise you'd create a supernatural "ghost" in the machine. I hope I don't have to explain why I consider creating ghosts unachievable. Even with photo or video AIs, there is no "new" content, just rehashed old content which is a subset of the trained data (why AI-generated photos often have this kind of "smooth" look to them). The only reason the output of AI has any meaning to us is because we give it meaning, not the computer.

So, before wasting millions of compute hours on this project, I'd first try to hire and indebted millennial who will be glad to finally put his philosophy degree to good use.

fschuett··on Show HN: Adventures in OCR
> After these experiments, it's clear some human review is needed for the text, including spelling fixes and footnote placement.

I just use ChatGPT for spelling fixes (i.e. when rewriting articles). You just have to instruct it to NOT auto-rephrase the article.

fschuett··on ChatGPT Pro
If any company wants me to be interviewed by AI to represent the client, I'll consider it ethical to let an AI represent me. Then AIs can interview AIs, maybe that'll get me the job. I have strong flashbacks to the movie "Surrogates" for some reason.
fschuett··on Discord API Errors
I use Discord for notifications when someone starts a "Chat with customer support" chat on my site. I tried various "chat button" solutions, but I ended up writing the chat system myself because no solution was really tailored to my task (esp. custom file type uploads and Google Ads integration, so I know which ad started a chat). I need to react 24/7 to new customers, so I needed a tool to ping me with notifications on my phone. The chat messages are still saved in the company database, but I send a link with the Chat ID to Discord, so that someone can instantly login to the admin panel and reply to the customer.

Using Discord, all I had to do was to create a new, empty company server, copy the webhook URL and then POST to the URL and boom I get a ping notification with the Chat ID on my phone (or I can @ a colleague to distribute the customer support for different timezones).

It's not a very "orthodox" solution, but it certainly sucks less than the alternatives (plus it's free).

fschuett··on Leaving and Waving
Remember friend, as you pass by

As you are now, so once was I

As I am now, soon you may be

Prepare for death and follow me

fschuett··on Ask HN: What is the most useless project you have worked on?
I worked on a tool to digitize the German land registry, only to later get notified that they already had such a tool, so my work was completely useless. Worked on that for about 8 months, at least I got a bit of payment because I did it on the job. Currently the digitization of the German land registry is done in PDF format, which is of course horrible and my taxes are right now paying for people manually copying and pasting text from PDF documents.

I was employed as a simple office clerk (I had failed to start my own startup and needed any job to get some money at the time) and was told to do the same as everyone else, copying text from the PDF, rewriting the text according to legal guidelines and pasting it into an ArcGIS-based software. Meanwhile the ArcGIS-based software I was supposed to paste into crashed every 30 minutes. So the only way I saw to actually do my job was to write a digitization tool for the PDF documents. Since I had internal access to government documents (which aren't publicly available to anyone of course) I saw it as a way to build my portfolio and put the project on my resume to find a better job afterwards (which I did). I had a good manager who allowed me sudo access (my state used Linux, yay!), so I got to work.

After about 6 months I had a full GUI tool written and digitized about 100 PDF files. The software can digitize from PDF to JSON, then ran a python.wasm VM over the digitized JSON in order to rewrite the texts according to the (constantly changing) legal guidelines. Then I also "hacked" the ArcGIS-based software with a HTTP middleman logging server and found out that it was just submitting XML requests to an endpoint, so I wrote a tool to batch upload the JSON documents directly into the database instead of having to launch ArcGIS. In the end I also wrote a management server that used libgit2 to create diffs between legal document (i.e. creating Grundbuchänderungsmitteilungen via git diff), I am very proud of this because it's cool, although the chance of ever getting official approval was almost zero (but hey I thought it was cool).

I was hoping I could somehow sell my tool and support for it to the government, since the project of land registry digitization has currently taken over 20 years[1] and it's still not finished. Apparently the problem is that some bureaucrats want to make the absolute perfect data model and only release the software once the data model is perfect (which is basically never going to happen). My approach of making an extensible JSON-based model for now to just translate the PDFs - and then adapting it for use-cases later was rejected because, well, I am not a multi-million dollar company and I sadly failed to impress the government with my skills. Long story short, after about 8 months of work I was given rights to my program (even though I wrote it on government payroll, technically) but it was rejected. Oh well, at least I got a nice academic article[2] avoided shooting my brain out from boredom copying and pasting text and also advanced my career somewhat.

In case anyone needs such a tool, it's licensed GPL3 over at https://grundbuch-test.eu/

[1] https://www.grundbuch.eu/nachrichten/

[2] https://geodaesie.info/zfv/zfv-archiv/zfv-147-jahrgang/zfv-2... (page 4)

fschuett··on Bun v1.0.0
Since it's a bundler, I guess something like "bun deploy" to make money from hosting JS apps. It's a common strategy, getting mass developer adoption first, then soft-pushing for their private hosting of applications.
← PreviousPage 3 of 4Next →