HNHacker News
TopNewBestAskShowJobs

ant6n

4,525 karma · joined September 15, 2014

founder luna-rail.com
submissionscomments
ant6n··on Using AI to write better code more slowly
There at a lot of crafts that don’t have real deep experts anymore because the work was 90% automated.
ant6n··on The bootstrapper's EU stack for under €10 per month
It genuinely is.
ant6n··on OpenAI Is Preparing to File for an IPO Soon
Tesla used to be a kind of “ethical” play. They made electric cars cool.
ant6n··on The last six months in LLMs in five minutes
I only used Claude first time in April, previously only ChatGPT and Gemini. And I struggle to see what the hype is all about - yes it seems a tiny bit smarter than the pack, but on the 20$ subscription it runs out of tokens in 5-20 minutes, and then you need to wait 3-4h.

ChatGPT 5.5 seems capable, although a bit stingy with “thinking” compared to earlier models, and I never run into session limits.

ant6n··on MacBook Neo Deep Dive: Benchmarks, Wafer Economics, and the 8GB Gamble
It’s a nice hardware. But that garden is full of bugs, not much less than if you have open windows.
ant6n··on If AI writes your code, why use Python?
That’s true. Once you have APIs and want to use classes to create larger structures, the language is full of warts.
ant6n··on I switched from Mac to a Lenovo Chromebook
...except the operating system. And the silly notch. And the weird keyboard. And the hard palm-cutting corner. And the reflective screen. and the finger-print-magnet materials. And the small amount of RAM. and the small SSD. And the weight.

Other than that, it's perfect! (On the blance,still better than any other laptop)

ant6n··on Wolfenstein 3D for Gameboy Color on custom cartridge (2016)
You need to do mid-frame tile updates just to show a full bitmap frame. There’s 360 8x8 tiles on the screen, but the tile indices are 8 bit (you can only reference 256 tiles). You can store only 384 tiles in VRAM - a bit more than a full screen. So the mid-screen update is to go from one tile dictionary to the other, so you can access 360 tiles in total.

You can update 1 tile per scan line (during hblank), so 154 tiles per frame (including 10 vblank scanlines). So you need 2.5 frames to replace all tiles.

If you are really smart about updates, you can “race the beam”, basically start updating tiles just as the frame starts rendering, just behind the active scan line. Then you can update maybe 280 tiles before the active scan line of the next frame catches up with you.

ant6n··on Neanderthals ran 'fat factories' 125,000 years ago (2025)
…so you were produced in Neanderthal fat factory?
ant6n··on Neanderthals ran 'fat factories' 125,000 years ago (2025)
Soap, soup, goop
ant6n··on Neanderthals ran 'fat factories' 125,000 years ago (2025)
Isn’t that what George Lukas said about Star Wars?
ant6n··on Kimi K2.6 just beat Claude, GPT-5.5, and Gemini in a coding challenge
What I would like to see is a comparison of how well the models work in long running conversations:

  * do they lie and gaslight

  *  do they start breaking down on very long chats (forget old context, just get dumber)

  * do they constantly try to tell me how smart I am vs solving the problem (yes man)

  * do they follow conventions, parameters set out early in the prompts, or forget them

  * if they cant read a given file (like pdf), do they lie about it

  * is there a branch function to go back to earlier state of conversation

  * what is the quality of the presentation of results (structure, wording, excessive use of tables, appropriate use of headings)

  * how does the bot deal with user frustration (empathy?)
For example Chatpgt 5.5 is fairly smart, but presentation of results is kind of poor and unstructured, and unnecessarily long. It will break down on long conversations (the long answers dont help here), and it can’t deal with that except lying and gaslighting. It also has very little empathy, and mostly ignores user frustration. But at least theres branching, so one can go back without completely starting over.

Gemini doesnt feel quite as smart these days. It does well with very long conversations. Except it has bugs where all context gets lost or pruned, and it will lie and gaslight about it. Theres also no branching, so once context is lost you have to start over. Presentation is decent. Empathy is fairly good, except if users get frustrated, it gets more and more flustered and breaks down.

ant6n··on DeepSeek v4
It’s not just about “good” UX. It’s riddled with bugs.
ant6n··on AI uses less water than the public thinks
This sounds like being concerned about the adverse health effects of a steak due to sugar.
ant6n··on DeepSeek v4
Not if Gemini Lost all context already. Also, it doesn't really work well, a lot of the nuance and information simply gets lost.
ant6n··on DeepSeek v4
Funny how Gemini is theoretically the best -- but in practice all the bugs in the interface mean I don't want to use it anymore. The worst is it forgets context (and lies about it), but it's very unreliable at reading pdfs (and lies about it). There's also no branch, so once the context is lost/polluted, you have to start projects over and build up the context from scratch again.
ant6n··on GPT-5.5
My impression has been that ChatGPT-5.4 has been getting dumber and more exhausting in the last couple of weeks. Like it makes a lot of obvious mistakes, ignores (parts of) prompts. keeps forgetting important facts or requirement.

Maybe this is a crazy theory, but I sometimes feel like they gimp their existing models before a big release to you'll notice more of a "step".

ant6n··on 4-bit floating point FP4
That written document is prehistoric.
ant6n··on 4-bit floating point FP4
> In ancient times, floating point numbers were stored in 32 bits.

I thought in ancient times, floating point numbers used to be 80 bit. They lived in a funky mini stack on the coprocessor (x87). Then one day, somebody came along and standardized those 32 and 64 bit floats we still have today.

ant6n··on Anonymous request-token comparisons from Opus 4.6 and Opus 4.7
I thought it would be to get better, to stay competitive with the competitors and free models.
ant6n··on Filing the corners off my MacBooks
Well actually, your not well actually to the well actually was actually a well actually of the well actually. Just sayin’.
ant6n··on Improving my focus by giving up my big monitor
I dunno. I wish as a start, MacOS would improve window management, so I don’t keep jumping around the various desktops all the time and get lost.
ant6n··on A forecast of the fair market value of SpaceX's businesses
Going to orbit is actually useful already, cf starlink
ant6n··on Seeing like a spreadsheet
Biggest blocker: I can’t create reliable excel sheets that potential investors can look at in MS Excel, formulas tend to break.

If I can’t share the spreadsheet, it’s not very useful.

ant6n··on How and why to take a logarithm of an image [video]
The joke is that if you fill in the center, it shows the Droste effect of the image and kind of diminishes the magic of it.
ant6n··on We haven't seen the worst of what gambling and prediction markets will do
Wait, is the shooter in “insider” in this case?
ant6n··on Antimatter has been transported for the first time
How heavy is the micro black hole? How do you “drag” it?
ant6n··on Antimatter has been transported for the first time
Well, in a crunch I wouldn't like to be caught without a secondary backup.
ant6n··on EU Inc.: A new harmonised corporate legal regime
I was looking into using tokenize.it to get some of these benefits for a German GmbH. But I guess this eu inc will take years beige it exists.
ant6n··on Hollywood Enters Oscars Weekend in Existential Crisis
Projecting a 100 inch picture on a wall costs a fraction of OLED at the same size.
← PreviousPage 6 of 34Next →