boyter

It came to a head with that story about the kid making a clock and getting in trouble. Leo was all in on one side that it was about racism and such, with John suggesting he fell for a media lie.

I think John turned out to be right in the end. They did seem to patch things up with John appearing on a episode years later I think "a christmas miracle" was the title, but that was the last time he was on as far as I know.

When Leo cut off Dvroak and CJal I turned out. You really do need that sort of personality for people to bounce off otherwise its less entertaining.

Came here to post than and you already did. Thank you!

You are welcome. Glad to hear its working for you. I have a few ideas I am working on to improve its relevance too that I hope pan out.

Grep prints out every matching line. For some searches a LLM might do it will get a lot of noise, and it might have to make that search because it cannot be specific. Targeted search can reduce the number of tokens.

I suspect this comparison is against reading the whole codebase though compared to just getting the bits you need.

Interesting. I too have been working in this space, though I took a different approach. Rather than building an index, I worked on making a "smarter grep" by offering search over codebases (and any text content really) with ranking and some structural awareness of the code. Most of my time was spend dealing with performance, and as a result it runs extremely quickly.

I will have to add this as a comparison to https://github.com/boyter/cs and see what my LLMs prefer for the sort of questions I ask. It too ships with MCP, but does NOT build an index for its search. I am very curious to see how it would rank seeing as it does not do basic BM25 but a code semantic variant of it.

This seems to work better for the "how does auth work" style of queries, while cs does "authenticate --only-declarations" and then weighs results based on content of the files, IE where matches are, in code, comments and the overall complexity of the file.

Have starred and will be watching.

Been working on https://searchcode.com/ again which I bought back, albeit as code search tool for LLMs. It solves the “should I use this library” by allowing the LLM to inspect search and analyse it before integration. Can use it to compare multiple repositories before downloading. It comes with a large amount of token savings and can be really useful when wanting to learn about a codebase.

Since it does it anyway I added dossier pages to it as well https://searchcode.com/repo/github.com/rust-lang/rust Which is useful for humans, and shows what the system is creating.

Best part is that I get to use the tools I have built, so https://github.com/boyter/scc and https://github.com/boyter/cs to improve it which benefits anyone using those tools.

Many Muslims drink anyway. A lot of those in Iran brew wine/beer in their house.

Tobacco in Australia has been taxed to the point we have a huge black market for it now. You would have thought people would have learnt from prohibition.

You cannot police morals.

Id be ok with that if wine had the same taste. No alcohol free wine tastes even close, and none of them are good in their own right.

Some of the non-alcoholic beers are pretty good though and I am happy to drink them.

I reimagined https://searchcode.com/ since I realised LLMs have issues when it comes to understanding code you want to integrate. It’s useful for looking though any codebase, or multiple without having to clone it.

I use it when I have candidate libraries to solve a problem, or I just want to find out how things work. Most recently I pointed it at fzf and was able to pull the insensitive SIMD matching it uses and speed my own projects up.

I can’t find it right now, but there was a post about how ripgrep worked from a someone who walked through the code finding interesting patterns and doing a write up on it. With this I get it over any codebase I find interesting, or can even compare them.

I read this when it came out and having written similar things for searchcode.com (back when it was a spanning code search engine), and while interesting I have questions about,

    We routinely see rg invocations that take more than 15 seconds

Now I totally believe codebases like this exist, but surely they aren't that common? I could understand this is for a single customer though!

Where this does fall down though is having to maintain that index. That's actually why when I was working on my own local code search tool boyter/cs on github I also just brute forced it. No index no problems, and with desktop CPU's coming out with 200mb of cache these days it seems increasingly like a winning approach.

The disk cache has a huge impact. However they claim it’s for multiple searches so it should be in it.

Such a good read. I actually went back though it the other day to steal the searching for the least common byte idea out to speed up my search tool https://github.com/boyter/cs which when coupled with the simd upper lower search technique from fzf cut the wall clock runtime by a third.

There was this post from cursor https://cursor.com/blog/fast-regex-search today about building an index for agents due to them hitting a limit on ripgrep, but I’m not sure what codebase they are hitting that warrants it. Especially since they would have to be at 100-200 GB to be getting to 15s of runtime. Unless it’s all matches that is.

Its pretty easy to step over those limits.

Also localhost and presumably this are good for validating your logic before you throw in roles, network and everything else that can be an issue on AWS.

Confirm it runs in this, and 99% of the time the issue when you deploy is something in the AWS config, not your logic.

All of the above.

Stop reading the news. It makes you depressed or angry. Go hiking. Walk on the beach. Play with a dog or your children. Climb a tree.

Leave the slave slab phone at home, or delete every news and social app. Do not browse the web. Take a book and read.

It will be hard at first. Then it gets easier. Best thing I ever did.

Reminder. What passes for news today wouldn’t have registered for most people 100 years ago.

Chased the wrong thing. It’s the 1% lows that matter more generally.

If you have a high frame rate to start with it’s pretty nice and feels smoother. But a low frame rate turned into a high one looks good but feels laggy.

So arguably you never need frame gen for a game, since it only really works when it’s already pretty nice.

Not familiar with that tool. What follows is my best guess based on what I am seeing.

Serena looks to be a precision tool. Since it uses uses LSP its able to replicate a lot of what a IDE would allow and IDE for LLM's.

cs by contrast is more of a discovery tool. When you're trying to find where the work actually happens it can help you, and since there is no index involved you can get going instantly on any codebase while they are index.

You could use cs for instant to find where the complexity lies, and then use Serena to modify it.

Seems to be a load issue, hopefully easily resolved

    Request URL https://lightwave.so/api/register/ephemeral
    Request Method POST 
    Status Code 429 Too Many Requests

People who give away things like this tend to be good people. As such when someone comes asking for help or new things they are inclined to help.

Your response is where it should go when things get rude, but you don't want to start there.

I had never seen this before. Although I am now trying out lagrange and seeing what it can do.

Sorry for hijacking the thread on what is a great post, but is gemini common these days? This is literally the first time I have ever seen it, and it seems fairly interesting, albeit deliberately limited.

Annoyingly the learn more about it link https://gemini.circumlunar.space/ is now dead, as possibly might be protocol.

Perhaps adding some guide on how to hook this up to... well anything would be good :)

There is a lack of guidance for https://github.com/mark3labs/mcp-go/ which this is using as well so while everything is there, its hard to know how to make it do anything.

Ignoring the open-source vs free software discussions that are bound to come about from this well said. Large companies exploiting developers and abuse towards the maintainers is probably my biggest bugbear when it comes to this.

In fact I have a similar post https://boyter.org/posts/the-three-f-s-of-open-source/ which I redirect people towards if they become aggressive towards me when I am trying to help them. Thankfully I have only had to use it a handful of times.

Crawling, incidentally, I think is the biggest issue with making a new search engine these days. Websites flat out refuse to support any crawler [other] than Google, and Cloudflare and other protection services and CDN's flat out deny access to incumbents. It is not a level playing field.

I wrote the above some time ago. I think its even more true today. Its practically impossible to crawl the way the bigger players do and with the increased focus on legislation in this area its going to lock out smaller teams even faster.

The old web is dead really. There really needs to be a move to more independent websites. Thankfully we are starting to see more of this like the linked searchmysite discussed earlier today https://news.ycombinator.com/item?id=43467541

For CRUD I agree. A lot of what I am doing is a bit more complex then that.

I actually would be happy to just "vibe" code my way through most of the problems I deal with if LLM's were able to do it.

That said, they make a great intern or jnr developer you can hand tasks off. You have to review either way, but the LLM does it faster.

I detest writing CSS and HTML. I just find it boring fiddly and annoying. I have started doing "vibe" coding with LLM's. Giving a decent prompt produces results that are... pretty good.

Almost 100% in lighthouse for both mobile and desktop, responsive, reusable components, dark/light mode and a design that was better than I could do in the 2-3 hours I spent doing it (while sipping wine).

I know its not a solution for everyone, and probably won't work for the prettier designs out there, but you can go a long way with these tools this day.

I know there is a reluctance to not use LLM's for code tasks, and I am one of the largest critics, but for me this solves a real pain point. I don't want to write CSS/HTML anymore and these tools do a good enough job of it that I don't have to.

TIL had no idea code2vec existed at all. Thanks for pointing it out.

While old, the nice thing about word2vec models is how easily they can be imported and used by any language to create a vector search.

I have always wondered (but never done anything) about if you could train on code to achieve a similar result. I suspect there could be some value there even if its just to help identify similar snippets of code.

I doubt its the GC kicking in, but you could run it with the following environment variable set just to ensure that its not doing anything.

    GOGC=-1

    Summary
      ./mainnogc 1000 ran
        1.01 ± 0.06 times faster than ./maingc 1000

While you are right, in that you add lines to remove bugs, I suspect its still true to say given 2 projects, written in the same language by the same person, on average the one with more lines is more likely to contain more bugs.