Hacker News Reader

FiberBundleon AGI is not possible even in 10 years

This is inconsistent with people like Hassabis or Sutskever giving time frames while also saying that LLMs won't get us to AGI.

FiberBundleon My experience with Claude Code after two weeks of adventures

https://m.youtube.com/watch?v=XyQ4ZTS5dGw&pp=ygUZTWl0Y2hlbGw...

Not a "100x" boost, but a pretty good take on what tasks agents can do for even very good programmers.

FiberBundleon Gemini CLI

Do you honestly believe that the opt-out by Anthropic and Cursor means your code won't be used for training their models? Seems likely that they would rather just risk taking a massive fine for potentially solving software development than to let some competitor try it instead.

FiberBundleon SpaceX Starship 36 Anomaly

In the spacex subreddit there are comments claiming that key engineers have left the company because of differences with leadership/culture. Not sure how credible those are, but spacex has had suspiciously many failures recently.

FiberBundleon Ask HN: Anyone struggling to get value out of coding LLMs?

Where's the actual evidence for the productivity boost though? Wouldn't one expect a huge increase in valuable software products or a dramatic increase in open source contributions if llms provide this kind of productivity increase?

FiberBundleon Side projects I've built since 2009

Would you be ok with giving a range of selling prices?

FiberBundleon Building an AI that watches rugby

I'm not a lawyer, but my interpretation of the lawsuit based on the Wikipedia article is that game results/scores are public facts and hence not copyrightable data. I don't see how the method by which that public data is collected changes anything materially about that case. Are you saying that inferring the score based on the scoreboard is what makes this illegal (why?)? What if they would infer the score using motion/ball tracking instead?

FiberBundleon Building an AI that watches rugby

I'd say it isn't in the US though, at least there's precedent that makes me think that: https://en.m.wikipedia.org/wiki/National_Basketball_Ass%27n_...

FiberBundleon Hallucinations in code are the least dangerous form of LLM mistakes

You're an extreme outlier. Most programmers work with 1-3 codebases probably. Obviously you can't keep 800 codebases in your head, and you have to settle for your approach out of necessity. I find it hard to believe you get anywhere close to the productivity benefits of having a good mental map of a codebase with just good documentation and extensive test coverage. I don't have any data on this, but from experience I'd say that people who really know a codebase can be 10-50x as fast at fixing bugs than those with only a mediocre familiarity.

FiberBundleon Hallucinations in code are the least dangerous form of LLM mistakes

In my experience you just don't keep as good a map of the codebase in your head when you have LLMs write a large part of your codebase as when you write everything yourself. Having a really good map of the codebase in your head is what brings you large productivity boosts when maintaining the code. So while LLMs do give me a 20-30% productivity boost for the initial implementation, they bring huge disadvantages after that, and that's why I still mostly write code myself and use LLMs only as a stackoverflow alternative.

FiberBundleon Zelensky leaves White House after angry meeting

This has nothing to do with being left-leaning. If you disagree with HN's sentiment here, it just means as you have completely lost the plot.

FiberBundleon Who killed the rave?

A yeah the deer are just confused, no need to worry. Keep partying!

FiberBundleon Can AI do maths yet? Thoughts from a mathematician

It didn't appear in the last two years. We have had deep learning based autoregressive language models (like Word2Vec) for at least 10 years.

FiberBundleon OpenAI O3 breakthrough high score on ARC-AGI-PUB

This is actually the result that I find way more impressive. Elite mathematicians think these problems are challenging and thought they were years away from being solvable by AI.

FiberBundleon AWS and Azure Are at Least 4x–10x More Expensive Than Hetzner

https://youtu.be/-cEn_83zRFw?feature=shared&t=1282

FiberBundleon OpenAI co-founder John Schulman says he will leave and join rival Anthropic

Why can't you just debug this yourself? I don't think completely relying on LLMs for something like this will do you any good in the long run.

FiberBundleon Tesla's FSD – A Useless Technology Demo

There have been lawsuits. I'm unable to find the actual documents, but this article [0] reports on it and has some unbelievable quotes in it.

> Lin rejected Tesla's argument that LoSavio should have known earlier. "Although Tesla contends that it should have been obvious to LoSavio that his car needed lidar to self-drive and that his car did not have it, LoSavio plausibly alleges that he reasonably believed Tesla's claims that it could achieve self-driving with the car's existing hardware and that, if he diligently brought his car in for the required updates, the car would soon achieve the promised results," Lin wrote.

EDIT: found the court document; the quote is in the last paragraph on page 5: https://regmedia.co.uk/2024/05/16/teslaamendedcomplaint.pdf

[0] https://arstechnica.com/tech-policy/2024/05/tesla-must-face-...

FiberBundleon German public broadcasters open source their streaming platforms

Would agree with that, but just very briefly: the main characteristic that defines fascism is a strong cult of personality. The Afd doesn't have that at all. They also don't display any kind of expansionist ambitions, but are rather nationalistic in an isolationist kind of way similar to most of the right-wing populists who have gained popularity throughout Europe. I have very little doubts that some of the hardliners in the party and possibly even some of the more moderate people are authoritarian, but fascists they are not.

FiberBundleon German public broadcasters open source their streaming platforms

Doesn't seem as if the public broadcasters are doing such a fine job, if their viewers obviously don't even know the definition of a fascist.

FiberBundleon Tesla recalls all cybertrucks for faulty accelerator pedals

The shareholders would pay for Musk's pay package by having their shares diluted. It's not as if Tesla has to provide any money for that, so it's not really comparable to giving employees a raise.

FiberBundleon Tesla recalls all cybertrucks for faulty accelerator pedals

Maybe not let people work 12 hour shifts? This isn't the 19th century.

FiberBundleon Learning about distributed systems: where to start? (2020)

Would also recommend Kleppmann's book. After reading this you should be able to read some foundational papers. [1] is a good list for that.

I also learned a ton by reading the jepsen analyses [2]. [3] is also helpful, but was recently put "behind a paywall" via an O'Reilly online book, but the internet archive should still have all the content ([4]).

[1] https://dancres.github.io/Pages/ [2] https://jepsen.io/analyses [3] https://martinfowler.com/articles/patterns-of-distributed-sy... [4] https://web.archive.org/web/20230628001937/https://martinfow...

FiberBundleon Nix is a better Docker image builder than Docker's image builder

[1] also uses this approach.

[1] https://mitchellh.com/writing/nix-with-dockerfiles

FiberBundleon LLMs and Programming in the first days of 2024

> OP is clearly an experienced engineer

You think? He's the creator of Redis.

FiberBundleon Challenging projects every programmer should try (2019)

> Every CS graduate can implement Paxos

Is all of this satire?

FiberBundleon Maelstrom: A workbench for learning distributed systems

Slightly off-topic, but I have worked with another tool by the author (Jepsen), which is also written in Clojure and in fact a Clojure library. I'm fairly proficient with functional programming in general, but had never touched Clojure before. My experience working with Jepsen was quite frustrating simply because of the fact that it's a Clojure library. Clojure is dynamically typed and it happened fairly frequently that I had to dig into Jepsen's source code simply because some type errors were only caught very deep in the call stack. Jepsen itself is a really great tool and I'm a very big fan of it's creator (Kyle Kingsbury), but the fact that you can only use it through Clojure is kind of sad. I guess it gets better with more clojure experience, but a dynamically typed functional programming language is kind of too much of design flaw for me to want to invest more time in the language.

FiberBundleon Show HN: nbi.ai – Generative Business Intelligence

Do you use that data for training? Given this:

> We started as a freemium product and built a unique business data set. We used this dataset to launch NBI.AI - a Generative BI platform that can be connected to virtually any structured data source

it sounds as if you have done so in the past.

FiberBundleon Every database will become a vector database sooner or later

Personally I find such snark comments to be a really bad look from someone who apparently is associated with a product (judging from comment history).

FiberBundleon We are beginning to roll out new voice and image capabilities in ChatGPT

Isn't video prediction a substantially harder problem than text prediction? At least that was the case a couple of years ago with RNNs/LSTMs. Haven't kept up with the research, maybe there's been progress.

FiberBundleon What OpenAI really wants

How can one actually take anything Emily Bender and Timnit Gebru say seriously?