haolez

10 ms·

121.
by haolez 2y ago
Is it so much better? What's the difference? How does it do embeddings?
122.
by haolez 2y ago
But it should be easy to generate such data to train an AI to do that, if someone wants, no?
123.
by haolez 2y ago
And APL! It saves so much tokens :D
124.
by haolez 2y ago
Is this an active fork?
125.
by haolez 2y ago
Out of curiosity: is LuaJIT supporting only 5.1 because of technical difficulties imposed by newer versions, or is it simply that the design of newer Lua versions did not appeal to the LuaJIT maintainers?
126.
by haolez 2y ago
This only happens if people are evaluated based on OKRs, which is not what this tool was originally about. High Output Management, as far as I can remember, says that such a tool must be used solely for communication, i.e. to avoid having p
127.
by haolez 2y ago
"KR #3: the amount of bugs reported by users must not increase from our baseline" Solved :)
128.
by haolez 2y ago
I get this feeling as well, but then again debugging large SQL queries can be very opaque. Things like pl/pgsql help, but then it starts becoming more and more like a normal programming language.
129.
by haolez 2y ago
And it learns online.
130.
by haolez 2y ago
I think it's a little of both. Maybe generative AI algorithms won't overcome their initial limitations. But maybe we don't need to overcome them to transform society in a very significant way.
131.
by haolez 2y ago
Maybe creating your own AI agents with your own "touch". Devin, for example, is very dogmatic regarding pull requests and some process bureaucracy. Different tasks and companies might benefit from different agent styles and workfl
132.
by haolez 2y ago
Never had an issue with credit cards on AWS, Azure, GCP, Digital Ocean, Vultr and Scaleway.
133.
by haolez 2y ago
Same with me a few years ago. Their support told me that they didn't want me as a customer. It was my first interaction with them. I swear I'm just a regular nerd with a credit card :D They are bizarre.
134.
by haolez 2y ago
How can 360 peer performance reviews ever work? The incentives are against a fair evaluation: the reviewers have the incentive to overly criticize others so that they can stand out more. I'm not saying that everyone on a 360 review pro
135.
by haolez 2y ago
One doubt that I have about LLMs is what would happen if I could retrain the model on each interaction. Like, from scratch, without costs and instantly, with the whole dataset + today's interactions. Would this feel closer to (a dumb)
136.
by haolez 2y ago
I've accidentally made it work extremely well in the past with Discourse[0] (not Discord). It was meant to record decision making processes, but it ended up replacing almost all of our async communications. Messaging was used only for
137.
by haolez 2y ago
Asking it to be an asshole and to be opinionated. This seems to tap deeply into the LLM's knowledge.
138.
by haolez 2y ago
Did that a decade ago and it saved my marriage as well :D I'd lived alone before marrying and I did all my chores, but my tolerance for messy stuff until I got to action was way more than my wife's.
139.
by haolez 2y ago
If you want to scare people, do it while Incognito. And repeat the search bar suggestions while on it ;)
140.
by haolez 2y ago
Yeah, I know. But then you have to make sure that your IDE is using the correct environment, that the notebook is using the correct environment, that the debugger is using the correct environment. It's trivial to setup a venv, but some
141.
by haolez 2y ago
Sometimes I don't even use virtual envs when developing locally in Python. I just install everything that I need with pip --user and be done with it. Never had any conflicts with system packages whatsoever. If I somehow break my --user
142.
by haolez 2y ago
Or it will be the (huge) technical debt of tomorrow.
143.
by haolez 2y ago
This is just a hunch, but it feels like OpenAI has lost research power and is now focusing on engineering challenges, since that's what they can execute for now. The recent releases of o1 and Search doesn't seem to advance their c
144.
by haolez 2y ago
From the errata in the article: > RTP is commonly used as an abbreviation for the Real-time Transport Protocol, and is not descriptive enough. This is a protocol for reliably downloading large files. it is not designed to be a drop in r
145.
by haolez 2y ago
I've been having some difficulties with deprecated code and old patterns being suggested all the time. But I guess this is an easy issue to fix and will probably be fixed eventually.
146.
by haolez 2y ago
Well, most of their extensions to VSCode are proprietary. When their dominance in software development becomes irreversible, it's obvious that they will close things down and create new sources of income. The incentives are clear.
147.
by haolez 2y ago
How's the feature parity with Acme + plumber + other Plan9 stuff? This feels like a modernization of Acme and it sounds like a very good idea to me. If it's not immediately a better alternative to VSCode and other IDEs, it can be
148.
by haolez 2y ago
I had some success with simple aliasing at the beginning and end of chunks. In my next project, I'll try an idea that I saw somewhere: 1. do naive chunking like before 2. calculate the embeddings of each chunk 3. clusterize the chunks
149.
by haolez 2y ago
I have a doubt about vector searches. Does increasing the amount of dimensions always improve performance? Or is there an optimal (application dependant) point where, beyond that, performance starts to degrade? I'm asking this because
150.
by haolez 2y ago
It was a genuine question. I was really curious about other use cases besides the obvious one.

More ›