Juan

3.1K posts

Juan banner
Juan

Juan

@ditorodev

building better and fairer interviews at https://t.co/Eb1NuUwExM

Barcelona Katılım Eylül 2020
2.7K Takip Edilen359 Takipçiler
Juan
Juan@ditorodev·
@saltyAom This is amazing dude I’m so curious how you have adapted your workflow with AI, historically you have been like one of the most hands-on deep into trenches dev I have seen Pretty curious how you are doing to augment that with AI (assuming AI alone fails)
English
0
0
0
146
SaltyAom
SaltyAom@saltyAom·
I didn't expect that assigning at least one property in the constructor would spill the class property into Butterfly arena I swear handling memory management in JavaScript is sometimes harder than a system programming language due to how unpredictable it is
SaltyAom tweet media
English
6
0
71
5.1K
Sam Willis
Sam Willis@samwillis·
Electric Circuits is our new incremental query engine that allows any Postgres query to be incrementally updated and streamed to your application front end. It builds on DBSP for IVM and @DurableStreams to resilient transport. Incredibly scalable, and very, very fast 🔥
Valter Balegas@balegas

This is Electric Circuits: a new primitive that turns any static database query into a live one — fine-grained reactivity, end to end, from Postgres to your app. Built on DBSP and @DurableStreams

English
6
9
87
10.4K
Juan
Juan@ditorodev·
@atomic_chat_hq Laguna is the most impressive model launched this year after fable
English
0
0
0
48
atomic.chat
atomic.chat@atomic_chat_hq·
Laguna S 2.1 performs at GLM-5.2 level on building popular games with 6x fewer params! We gave three local models the same task: build three popular arcade games that play themselves. Each game is one self-contained HTML file with a bot that plays it. Prompts: – Geometry Dash – Doodle Jump – Air Hockey Outputs: Laguna S 2.1: 10.3K tokens GLM-5.2: 26.4K tokens Hy3: 10.4K tokens Laguna held its quality against a 753B model. We think Laguna's Geometry Dash looked the best of the three, the cube clears every spike and block. GLM won Air Hockey. Its table looked the most detailed of all. Hy3 was the only model that added shooting to its Doodle Jump. But Laguna is the only model in our benchmark that runs on a MacBook with 128GB!
Poolside@poolsideai

Today we're releasing Laguna S 2.1, our most capable model to date. It's a 118B total parameter Mixture-of-Experts model with 8B activated per token, a context window of up to 1M tokens, and thinking and no-thinking modes. Capable enough to hold its own against models many times its size. Small enough to run on a single @NVIDIAAI DGX Spark. Laguna S 2.1 is fully open under OpenMDW-1.1, with weights available today on @huggingface poolside.ai/blog/introduci…

English
21
46
604
66K
Juan
Juan@ditorodev·
agents.craft.do is an impressive harness, feels a lot like if pi + codex app had a child lots of things are just working for me, so far, for free i have: - remote server support - native claude-code like workflows - claude-code and pi available - ws + tg support
English
0
0
0
73
Juan
Juan@ditorodev·
@training_loop Yeah sadly im kinda very into fable 5 prchestration
English
0
0
0
29
tynan
tynan@training_loop·
@ditorodev the codex ux for this is so vastly superior
English
1
0
1
60
Juan
Juan@ditorodev·
lol Fable 5 moderation so stupid working on Auth for server to server protocols is flagged :D
English
1
0
0
92
Dillon Mulroy
Dillon Mulroy@dillon_mulroy·
what’s the best tooling to let agents use a browser rn? agents always seem to burn a ton of tokens using things like agent-browser esp using helium (@uwukko maybe you have some good guidance here?)
English
84
2
334
57.6K
Juan
Juan@ditorodev·
@Bharath_uwu as a famous guy said: "a 1.5 billion hole in the books is not my problem, thats now their problems, not mine"
English
0
0
1
48
Álvaro Monterrubio
Álvaro Monterrubio@alvaro_monsan·
Summer nights at @hirevoice building product and not fixing bugs (it finally works) 🏅
Álvaro Monterrubio tweet media
English
2
1
12
562
Juan
Juan@ditorodev·
@peer_rich is this on a repo somewhere? Curious if I can help, me mee de risa con esto "Ventanas" "Cromo" "Borde"
Juan tweet media
English
0
0
1
42
Juan retweetledi
boris
boris@boristane·
god tier ad
English
45
429
3.9K
189.8K
Juan
Juan@ditorodev·
@glcst This means I can cache my sqlalchemy queries on durable objects for free?
English
1
0
0
26
Glauber Costa
Glauber Costa@glcst·
What if we could run Postgres as a single file, and take advantage of the best SQLite has to offer? Today I am announcing pg-micro, a crazy experiment I've been undertaking to make this happen. pg-micro is different than other approaches because it is fully local, and expected to be fast: there is no concurrency limitation and no statement translation. Here's how it works: we use the actual postgres parser to parse the statement, but compile that to the Turso AST. The Turso AST is then compiled do bytecode, and from there everything executes natively, as it'd do in SQLite. This makes it a perfect target to run in any environment. There is traditionally a mismatch between Postgres and SQLite in terms of functionality. But @tursodatabase has been hard at work to close this gap: things like MVCC and a rich, strict type system are present in Turso. There are PRs for things like lateral joins, etc. This means that the gap can be closed until it theoretically reaches zero. What you could do with it? Just imagine for example a primitive like Durable Objects by @Cloudflare, but with a postgres interface? Or imagine you could use the same pattern of local databases for agents that SQLite gives you, totally ephemeral and free, but with a Postgres interface? Or even that you could execute remote postgres in platforms like @vercel but with the unmatched density of the Turso Cloud? Expect lots not to work at this point. But as usual, this is done in the full spirit of OSS, so PRs welcome! To get started: npx pg-micro
Glauber Costa tweet media
English
91
175
2K
211.1K
Juan
Juan@ditorodev·
im out of fable and that makes me sad
English
0
0
0
37
Juan
Juan@ditorodev·
one thing im seeing with ai is that most people dont do products for themselves because they get distracted and/or tired, still preferring to outsource it
English
0
0
0
47
Juan
Juan@ditorodev·
@stevibe Really love this
English
0
0
2
117
stevibe
stevibe@stevibe·
You know that "But, wait..." moment in every LLM thinking trace? I made it visible. I asked 8 models the same tricky probability question and rendered their reasoning as trees. Every time a model rejects its own idea and pivots, every "But...", every "Wait, actually...", a new branch grows. Same question. Completely different minds.
English
67
107
1.1K
86.5K
Juan
Juan@ditorodev·
a skill that forces the agent to write a custom workflow using flue framework seems to be a huge unlock in capability
English
0
0
0
22
Juan retweetledi
Gergely Orosz
Gergely Orosz@GergelyOrosz·
This is exactly why experienced software engineers are valuable and will be valuable. If you don’t know what good code looks like you will have no idea if what the models generate are any good Of course “AI reviews the code” etc etc… it doesn’t work as reliably. Via @mitchellh
Gergely Orosz tweet media
English
116
192
3.6K
240.6K