barrell

23 ms·

331.
by barrell 2y ago
So when you click on a lesson - you like the variance of all the review types before the lesson finishes? That's very doable! Definitely somewhere I'm working towards, and I have a loooot of review ideas XD it'll be a few mon
332.
by barrell 2y ago
If you were in charge of Lingvist, what would you change/build? edit: long time lingvist user, I'm also a huuuuge fan.
333.
by barrell 2y ago
Could you elaborate on which features you'd miss from Duolingo if you switched? For context I'm working on a language learning app (phrasing.app) and I'm pretty against gamification. I've got a couple different review me
334.
by barrell 2y ago
Phrasing ( https://phrasing.app ) | CTO | Europe (REMOTE) We're looking to assemble a small but mighty team to topple giants. Together, we'll leverage AI, ML, and LLMs to build the world's most effective language le
335.
by barrell 2y ago
For me, a couple things; - fast enough to compete with neovim. Idk if it’s my previous interest in display engineering, but I substantially notice the speed - vim bindings…. Satisfactory. I don’t struggle to navigate at all, feels pretty na
336.
by barrell 2y ago
Phrasing | Full-time | CTO | REMOTE (EU) We're looking to build our own small, cracked team to take on Duolingo and bring language acquisition to the masses. The team is currently 1 engineer and 1 designer. Tech stack: - ClojureScript
337.
by barrell 2y ago
No it's not a problem with the underlying infrastructure. I believe the OP was asking why use this product, not why is this specific infrastructure necessary. The infrastructure before was working fine (with SAQ at least, Celery was an
338.
by barrell 2y ago
I have a bunch of different queues that used SAQ (~50 LoC for the whole setup) and deployed it to production. A lot of them use LLMs, and when one of them failed it was near impossible to debug. Every workflow has over a dozen connected tas
339.
by barrell 2y ago
I don't have any experience with prefect, but I have to say one of my favorite things about SAQ (Simple Async Queue) was a task was a task was a task. You could enqueue them from anywhere, nest them, repeat them, skip them, whichever.
340.
by barrell 2y ago
I’ve been through a whole journey with distributed tasks queues - from celery, to arq, to recently hatchet. Not only is hatchet the only solution that doesn’t make me want to tear my hair out, but the visibility the product gives you is ama
341.
by barrell 2y ago
Also IIRC (and cursory searches seem to agree) blue is the most common/prominent color other animals see
342.
by barrell 2y ago
Funny, I have the complete opposite experience with python. Constantly turned off inline errors and warnings because they were almost always wrong - packages I installed “could not be found” by the LSP, it constantly worried about type issu
343.
by barrell 2y ago
Only if you’re comfortable being converted to lisp
344.
by barrell 2y ago
Came here to say the same thing, it sums it up perfectly
345.
by barrell 2y ago
Fixed! Didn’t know we couple update posts after 24 hours PS thanks ;)
346.
by barrell 2y ago
From what I gather it’s a combination of immutable data structures, light caching, and running the lsp/treesitter in background processes. The founder has a bunch of interesting interviews going deep into the mechanics if you’re intere
347.
by barrell 2y ago
Zed has been amazing so far for my Elixir journey, and their latest preview release has additional support for Elixir tests. Not sure how it compares to IntelliJ
348.
by barrell 2y ago
Can confirm, am early stage founder who is a true believer that BEAM is some kind of magic scaling bullet or secret sauce, currently looking for elixir dev. (mostly joking, if anyone is actually interested in ML, NLP, and Elixir, I have mor
349.
by barrell 2y ago
Which parts of his application need a 10x speed improvement?
350.
by barrell 2y ago
I get this. Currently have a front-end built, with the need (& existing functionality) for offline local first etc. Switched to phoenix on the backend, and the amount of liveview stuff I have to 'wade through' is really distra
351.
by barrell 2y ago
Recently moved my python app over to Elixir, and currently gushing over Elixir every day. I'm a huge fan and have loved everything about it, except... Comping from a LISP, the Elixir REPL really is not world-class. It's a really f
352.
by barrell 2y ago
Don't forget: - ExUnit for incredibly easy tests - Hex package manager that just works - FLAME for the ability to scale different processes to different computers with nearly 1 line of code (the whole library itself is just a few hundr
353.
by barrell 2y ago
- They outperform transformers at lower parameter counts. Time will tell if that hold up for more parameters - They scale linearly in complexity, which means with a longer context window they will be faster and cheaper than transformers - I
354.
by barrell 2y ago
Phrasing.app | CTO | Remote (Europe) | Full-time | Amsterdam based | Clojure, Elixir, ML, Swift We're a small scrappy team of two building a language learning app to rival the giants. We've just launched privately, and are getting
355.
by barrell 2y ago
I would also guess that Elixir led to some crazy things they didn’t even imagine possible with web, like phoenixs live view [1]. Even if they don’t have explicit ideas of how it will impact ML, it’ll be really interesting to try. [1]: http
356.
by barrell 2y ago
I’ve been considering using elixir for a while. With these libraries, it’s now entirely feasible to move my backend over to elixir, getting rid of 95% of my python code in the process. The tasks leftover that still need python can easily ea
357.
by barrell 2y ago
Even with all those comments, it still ends up being less characters than other languages :D
358.
by barrell 2y ago
In this comment I was referring to POS tagging and feature extraction. Another use case for fine tuning we have as well is to reduce a 5-shot prompt that we have to run hundreds of times per request down to a “0-shot” (heavy emphasis on the
359.
by barrell 2y ago
When it comes to traditional NLP related tasks, LLMs are far below dedicated NLP pipelines like POS tagging and feature tagging. However, fine tuning bridges the gap quite a bit between the two. It's a narrow domain, but so is most of
360.
by barrell 2y ago
Without commenting on the debate at large, it’s a bit funny to read this comment. I mean voice cloning a year or two ago was basically science fiction, now we’re talking about voices being distinguishable as proof it’s not cloned, sourced,

More ›