senko

37 ms·

811.
by senko 3y ago
The problem is not (only) the context size, it's (for lack of a better word) focus. GPT4 can easily get lost in too much information and produce results that don't work well (duplicate code or just incoherently solving a problem),
812.
by senko 3y ago
> I'd really like it if Cognition Labs would put the resulting code from the demo into an open-source repository so we could examine it directly. Yup. It's hard to evaluate things based on the demo. We're building somethi
813.
by senko 3y ago
I would clarify that "there" in my "not there yet" doesn't assume superhuman AGI developer that will automagically solve all the software development projects. That's a deep philosophical issue best addressed i
814.
by senko 3y ago
In my experiments at Pythagora[0], we've found that sweet spot is technical person who doesn't want to know, doesn't know, or doesn't care about the details, but is still technical enough to be able to guide the AI. Also
815.
by senko 3y ago
I am making a sales pitch on behalf of all the projects I mentioned (not just the one I'm involved with). I see LLMs failing at coding daily (one of the "perks" of working in the space), and I'm incredibly bullish on t
816.
by senko 3y ago
As someone who works in this space ( https://pythagora.ai ), I welcome new entrants to this niche. Currently, mainstream AI usage in coding is at the level of assistants and glorified autocomplete. Which is great (I use GitHub Cop
817.
Detecting AI-generated job applications using prompt injection(blog.pythagora.ai)
4 points by senko 3y ago|0 comments
818.
by senko 3y ago
If your friend wasn't trolling and genuinely believed this advice, they have a very poor understanding of risk, and it would be beneficial to them to research it a bit (by which I don't mean click-driven youtube or tiktok influenc
819.
by senko 3y ago
A daily digest email for HN, and ArsTechnica (two separate projects, same idea). It gathers top 50 stories from HN (and all articles in case of Ars) for the previous day and sends me an email with the links (along the number of points/
820.
by senko 3y ago
The problem you have is with the record labels, not Spotify. Spotify takes the blame, but the record labels take the lion's share.
821.
by senko 3y ago
I haven't, but tbh we work a lot more with GPT than Claude so it's possible I haven't encountered many warts there. For what we do (AI code writing), GPT output seems qualitatively much better than Claude's, but we want
822.
by senko 3y ago
We've recently tested long context recall across Claude (2 and Instant) and GPT (3.5 and 4), results in https://dev.to/zvone187/gpt-4-vs-claude-2-context-recall-ana... Claude2 beats GPT4 in recall reliability, but
823.
Proof of concept for AI dev tool to create apps from scratch under supervision(blog.pythagora.ai)
2 points by senko 3y ago|0 comments
824.
by senko 3y ago
Remember, this is the project that raised ~$30m from Benchmark and Sequoia. There was a controversial "quality doesn't matter for software products" post and discussion[0] here on HN a few days ago and this is a beautiful exa
825.
by senko 3y ago
> Semver is the tool, not the cause. > If product maintainers do not care about stable versions, it makes no difference if they comply with semver or not. Exactly. Here, the maintainer reserves the freedom to break anything at any mom
826.
by senko 3y ago
This is semver: > Major version zero (0.y.z) is for initial development. Anything MAY change at any time. The public API SHOULD NOT be considered stable. (from https://semver.org/ )
827.
by senko 3y ago
With the latest 3.5-turbo, you can try forcing it to call your function with a well-defined schema for arguments. If the structure is not overly complex, this should work.
828.
by senko 3y ago
"Software quality" is such ambigously defined term that all of these can simultaneously be true (for different products): * software quality is (more) important * software quality is not (very) important * bugs are not important (
829.
by senko 3y ago
Yes. Open source is supposed to be merit-driven, not hype-driven and resorting to dirty tricks like these. I wouldn't single out the JS community is the only one affected by it. To name just two examples outside JS, I blame pipenv auth
830.
by senko 3y ago
I write very infrequently, but here it is: https://blog.senko.net
831.
by senko 3y ago
> It is good to know, as an author, that you should not accept releasing stuff under a license that has an "or later" clause. Definitely: https://opensource.stackexchange.com/questions/4077/can-the-...
832.
by senko 3y ago
Not parent, but thank you for the thoughtful explanation. Which free/open source licences do you think would be better for the specific purpose of preventing SaaSification while actually being usable/useful in real life?
833.
Fiscal Dominance (Lyn Alden July '23 Newsletter)(lynalden.com)
2 points by senko 3y ago|0 comments
834.
by senko 3y ago
If you have React engineers, why couldn't you have AI engineers? I think the post is right to try to delineate the roles of "builds LLMs or other ML models" and "uses those models", as (as you noted) the skills requ
835.
by senko 3y ago
Oh I'm sure I can't recall a bunch of (details of) books I've read. On the flip side, I have a firm belief that every book worth reading changes you at least a little, so even if I don't remember it consciously, it has h
836.
by senko 3y ago
The internet is a mob and the current social networks make taking out pitchforks as easy as clicking a button, and are happy to amplify the outrage.
837.
by senko 3y ago
Why would you need a list at all? If it's worth remembering, you'll remember you've read it, and if you want to refresh your memory what better than to reread?
838.
by senko 3y ago
Since you completely ignored the parent's valid point (many things that are in fact healthy for you can also cause cancer if overdoing, so the phrase "may cause cancer" by itself is not a valid argument), here's why your
839.
by senko 3y ago
Unnecessary ad hominem. Every research tank is subsidized by someone .
840.
by senko 3y ago
LLMs are the machine-translation services created specifically for that purpose[0], it just turned out they're very good at many other things! Your analogy would be like saying why use a computer to multiply numbers if you can calcul

More ›