Hacker News Reader

weiliddaton Compression is prediction

Relevant old school compression benchmarks where people have been using different models (incl. transformers) for compression: https://www.mattmahoney.net/dc/text.html

Also interesting the top entry is from fabrice bellard: https://bellard.org/nncp/nncp.pdf

weiliddaton Itadakimasu: A word you say to the food, not the cook

I think someone else wrote it better than I can: https://news.ycombinator.com/item?id=49080703

This is worthy feedback to the author (and others), that writing with clear AI tells/register is so distracting that it takes away from the spirit and content of the writing.

Perhaps this is just the zeitgeist and in 20 years this is irrelevant. It’s relevant for whoever is writing now.

weiliddaton Itadakimasu: A word you say to the food, not the cook

Why?

weiliddaton Itadakimasu: A word you say to the food, not the cook

https://www.pangram.com/history/c837bbcb-4aed-43e1-987c-afe2...

I know you're being downvoted, because people are tired of hearing "this is AI".

But I think your comment "this is AI" is warranted. This content, explaining something of cultural, personal significance (instead of the typical technical explainer) should really be written by a human.

weiliddaton The Economic Benefit of Refactoring

I'm sure you're where you are in some part due to hard work, skill, decisions. But surely you don't believe that it's purely down to skill where one lives/works at?

weiliddaton The Economic Benefit of Refactoring

Don't disagree that that's the system not working well.

You're lucky to be in a system that works out for you. Not everyone is in such a lucky position. LLMs help automate the tedium out of their job, and hopefully, have a bit more extra energy to make a better system within their immediate locus of control.

weiliddaton I flagged two research papers for fake authors and both were accepted as orals

Yeah exactly. One more metric isn't going to solve it.

weiliddaton DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

I don't necessarily mean reguritating it, but choosing which part of the scripture to surface to the user is already some interpretation/choice. Even the devil can quote scripture (I'm playing the devil's advocate here).

weiliddaton DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

Yes it can point you to the right place, but choosing which part of scripture to point you to is already a choice/interpretation

weiliddaton Agent Skill to Force Docs in ASD-STE100 Simplified Technical English

Anecdote: was testing Opus 5 and even ASD-STE100 cannot save you from its meta-commentary and signposting all the time. It writes what it wants to write, e.g. "here's where you are right/wrong" instead of "Your instinct of the shape is right, but here's a sharper version". I could not get it to stop doing that.

In contrast if you ask GPT 5.5/5.6 to speak directly/plainly, it will just straight up tell you the content "X is right, Y is wrong". Much preferable.

weiliddaton The Economic Benefit of Refactoring

> They'd be 10x better than us already if task tedium was the problem... the areas where people think that LLMs excel at coding and don't like doing manually it's usually because the human was inclined to slop out repetitive boilerplate and thought that was the only way.

There is so much work out there that is repetitive / boilerplate / tedium. If you get to personally work on interesting work more than 50% of the time (pre-LLM) I'd say your job is #blessed.

weiliddaton DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis

I get where you're coming from, and the intent to make it easier for people to find examples and verses, but there's a fine line with LLMs giving you answers, is that it's interpreting it in some form. Doesn't that run counter to prevailing ideology, that you're meant to either struggle with the materials / seek understanding yourself, or have your religious leaders interpret/receive those insights?

weiliddaton The AI Aesthetic

Kinda, more accurately the current goal for labs is to produce economic value to justify high valuations / capex, thus having to RL on correct answers rather than interesting ones.

If you could access the current models as base models I guess you'd still get very interesting hilarious stuff.

weiliddaton Advancing the price-performance frontier with GPT‑5.6

idk if you're still actually looking for a solution (I got nerdsniped heh), but I found a seamless way to have it locked into the speakers + room, instead of a specific source/software, is using an iLoud subwoofer that routes to any speaker setup and has room correction + EQ that lives on the sub.

weiliddaton The AI Aesthetic

What you said made me think that it's similar (not entirely the same, but adjacent) to those make-a-website for $XX or buy-a-template era of websites.

If we think about that as the intended competition (and not websites corps spent millions on design/polish/copy), I'd argue that AI (model + coding agent + skills + whatever) is doing a better job overall. Most agencies don't spend enough effort in understanding what you want, or hire juniors and bill them as seniors. It's almost always late, overbilled, somewhat what you said you wanted but not really, and technically terrible.

weiliddaton The Religion of Speed

I like this argument, but I don't think the article argues against taking urgency into consideration. Do you have anecdotes about companies or people that actually ignore urgency?

I've experienced enough business-critical incidents (or requests) where people across all seniority are just randomly trying stuff, making a mess, without taking time to coordinate, understand the problem, and solve it. It almost always resulted in more overall time and bigger blast radius than if we did it not out of panick.

OTOH my "slow" (I still did things fast, just not purely for the sake of looking fast) and steady approach always took me less time overall to find a proper fix. I also became known as the person who could always find the best possible fix when there's an incident/time constraint, but I also thought that was systematically bad instead of fixing the culture/process.

It's still faster to take a minute (and deep breath), understand the whole problem (incl. urgency and how that affects your solutions), and then solve it. It doesn't mean you can't keep your customers in the loop and assuage their concerns.

weiliddaton I flagged two research papers for fake authors and both were accepted as orals

Yeah it's a well-known secret, and I think any sort of scoring system misses the point. The scentific method (and system, incl. peer reviews) is a public good (like good governance) that requires moral acknowledgement, the necessary social/cultural/political/academic pressures, and discipline of taking care of it by participating parties.

If you add a score/metric this incentivizes the wrong thing, like how money, funding, and career already incentivizes all the wrong non-scientific behaviors (like faking data, p-hacking, etc.)

weiliddaton Teach yourself programming in ten years (1998)

Hm since the discourse is mostly on how AI changes this, I still want to make a point that it's worth learning programming (and more), even if I'm a huge proponent of AI being economically valuable.

Humans at the moment are still better at building on our own knowledge and compounding our effort:output ratio, due to our superior plasticity. See https://joyemang33.github.io/blog/2026/humans-dont-just-samp... for a small experiment. That's why for programmers experienced with a particular language/codebase/domain, it feels like minimal effort to know where and what to change. "Why use LLMs when I can do it in 10 minutes"?

Of course, relying on our long term memory instead of re-ingesting the code, is also a problem for having assumptions that don't get updated, e.g. revisiting a codebase that's been updated by many others. With the rate of change/churn of codebases I'm wondering if this is naturally forcing us to use AI anyways, since it's really hard to keep track of a lot of code changes by AI as a human. Once you start down the path of not reading code yourself, the friction to start reading code gets higher in a feedback loop. I know people (incl. myself) who are still reading code in their own coding/agent sessions, but that's unsustainable in an organizational setting - you won't be able to ingest N'0,000s of LoC of changes every week if there is no incentive to do so.

Is that a good thing? It's unlikely LLMs in their current state will be "neuroplastic" for a single person's use case (mostly market reasons - cost of owning, running, tuning models will likely be unattainable for the foreseeable future). In a few generations, this would be a total surrender of all foundational coding to LLMs (unsure about knowledge work in other domains). Will text files with superintelligent models be enough to encode long term and broad understanding of a domain? Is it necessary if a model is superintelligent? What if a person/nation/species loses access to LLMs through supernatural / political reasons at that point? Will it be like losing the last remaining COBOL programmers to your mainframe? I'm sure we're resilient enough to re-learn all of that, but it will be a huge crutch like never before.

All that to say, it's still worth learning programming, and doing some things by hand, even if it's not "economically valuable" (at least on the surface). Even with web frameworks, it's still worth manually plumbing between frontend/backend/database. Even with interpreters and compilers, it is still worth understanding and doing a bit of assembly. Even with assemblers, it's still worth understanding machine code, CPU and architecture quirks. It's worth looking at the historical artifacts (both senses of the word) and understanding intents, constraints, and creativity (as another comment [1] has more beautifully said).

Two "further than programming" courses that also embody the spirit of the article (get interested, hands-on, play) have been impactful on my appreciation of computing systems are nand2tetris [2] and OS in 1000 lines [3]. Both are "easy" in the sense that they make certain design decisions so you don't have to dig through thousands of pages of manuals. If you wish, during or after the process, you can explore specific concepts further. It's broken down in the right pieces and checkpoints, so you get quick feedback on whether something is working or not. It's still very much a puzzle you have to solve, but the right difficulty/complexity ladder (great instructional design).

[1] https://news.ycombinator.com/item?id=49093468 [2] https://www.nand2tetris.org/ [3] https://operating-system-in-1000-lines.vercel.app/en/

weiliddaton An Uncomplicated Man – A review of The Odyssey, directed by Christopher Nolan

Orthogonal to the movie/review, but if I could upvote this 100 times I would. These 3 questions were drilled to me in an elective Study of Visual Arts course in college and changed how I thought about everything and everyone.

weiliddaton An Uncomplicated Man – A review of The Odyssey, directed by Christopher Nolan

I read it more as a lament than a real critique. Here's all the ways it was different than the source material.

I think that was Nolan's intent all along - a very Nolan, American/Hollywood take on the Odyssey.

weiliddaton Benchmarking Opus 5 on SlopCodeBench

For most of these benchmarks, I feel like a p50 and p95 (using SWE salary as a proxy?) human benchmark as reference would be interesting.

Edit: FWIW the paper the post quoted has repositories as slop baseline https://arxiv.org/html/2603.24755v1#S4.SS2

weiliddaton Octane, the successor to Inferno – React's programming model, compiled

I couldn't access the post for some reason. Docs: https://octanejs.dev/

weiliddaton Qwen 3.8

I read his followup tweet, and your comment, and I'm not fully convinced that open models are decelerationist. Happy to hear other thoughts on this.

Open weight AI is decelerationist from the perspective that all capital should be allocated to a market leaders for training, and that the market leader is fully invested in continuously making the models smarter, cheaper, faster for its users, or that distillation from this market leader is the main way to make progress.

We might reach a local optimum/equilibrium faster without open weight models, with leaders capturing more of the market faster to a point where further R&D isn't required due to lack of competition. I also doubt that distillation is the only/main way that open weight models were advancing AI research. We can name a few examples from DeepSeek around reasoning, context optimization, etc. I'm also unconvinced that the overall market capex on AI is lower given more competition (probably less specifically for US market capex, which is decelerationist from only the US perspective).

weiliddaton Fable 5 vs. GPT-5.6 Sol on an NP-Hard Problem: Does /goal help?

One interesting harness thing I saw Cursor do is to give the model access to the entire thread.

Even if it doesn't fit in the context window, the model can search through past turns and sanity check if something doesn't seem to be going right, or be prompted to follow an early message, "when starting on a new item, review the first message for how we should approach this"

weiliddaton What ORMs have taught me: just learn SQL (2014)

I was curious how have sentiments changed over time. Brief LLM-based analysis: https://ampcode.com/threads/T-019f32ac-3b1e-74be-ad63-5f175d...

Overall, seems like it got more nuanced over time - even though it's still broadly in favor of SQL. Favor for ORMs (flagged also as a term that can mean many things to different people) is more in terms of type safety, mapping, migrations, etc. so more a library/utility rather than a framework that fully abstracts away the database.

weiliddaton The Coming Loop

> Often, it takes 5-6 broken crappy versions of a thing until you understand that. There is no accelerating the 5-6 broken crappy versions - there’s no agent tech that’s going to help your meat brain avoid thinking time.

Fully agreed. Though I found that, once I found a harness (prompt + skills + model) that I trust to do most things the way I like it, it has sped up the coding/exploration part of that process.

Although the iterations are faster, it's still taking me almost the same amount of effort to go through those crappy versions, because I still need to understand and adjust my mental model of what ideas, principles, design apply for the solution, and what to try next.

So, in the end, I do feel like I'm expending more mental effort in a shorter amount of time (with some effort saved on writing the code, which wasn't that much to begin with once you're proficient). It's a weird feeling, like I'm "only" prompting and reading code, but I feel equally mentally spent, or sometimes more because of the compressed iteration cycle.

weiliddaton Please Do Not Vibe Fuck Up This Software

No harm done, glad I clarified.

I'm generally an optimistic person and very trusting of others. I'd say I'm also a pretty good reader of intentions / listener based on people who are my friends / worked with me (anecdotal of course, take it how you will).

However, some of the comments in the GitHub issue... I can only assume the worst of intentions to ruin any/all motivation of the maintainer. Given we've seen social engineering and other attacks on other open source projects with increasing frequency, I can only assume that there are ulterior motives in such comments.

I cannot otherwise see how those comments would be constructive towards the maintainer or the other participants in the issue.

weiliddaton Dav2d

Supported is different from doing it well though. You do notice the performance hit even on TVs that playback YouTube videos on AV1.

Even on 1080p videos running on AV1 on 1x, the TV system bogs down and any kind of interaction has a variable 1-3s lag. On some TVs if you do 1.25x the TV automatically "downgrades" the resolution to 480p to avoid dropping frames.

I wish there was an option to still use VP9 / H.264 on those systems (even limited to 1080p).

weiliddaton Please Do Not Vibe Fuck Up This Software

I'm not sure how to interpret your comment. It could be

- a response to my comment saying that I am "illiterate" and cannot differentiate LLM output vs actual human comments (in that case I'm not sure what you're adding to the discussion here beyond a personal attack)

- a general comment saying it's getting harder for people in a position similar to us (i.e. tech / tech-adjacent who interact a lot with others who write with LLM assistance or via LLMs) to differentiate human/AI output.

I'll assume good faith and you mean the second. In that case maybe you can explain the "fundamental problem" you're referring to?

weiliddaton Please Do Not Vibe Fuck Up This Software

Maybe I'm getting too skeptical. I have a feeling increasingly many of the comments on HN and the GitHub issue are just bots ragebaiting other people (incl. the maintainer)...