Supreme Leader Wiggum

If I were to stream… and it was mostly just me doing my day to day work, while talking to chat, would anyone watch? Setup isn’t going to be anything stellar, no real agenda - I’d pretty much just go do stuff across 5 or 6 repos in parallel & talk through what I’m thinking

No. It converts posts and content into numerical representations for various algorithms which in turn are used to train model weights. Like novelty, expected decay, industry. I use embedders for creating clusters and vectors on the actual original text, but that’s not useing the stuff the LLM does.

@ScriptedAlchemy So it processes this and determines picks, entries and exits?

Reading 5000 news articles is not actually that expensive. Hell, I had single threads that totaled 300 million tokens this afternoon alone. I need to ingest a total of 24,000 events. (Bluesky posts and news feeds) - the LLM converts this into a json object with various fields.

@thekitze @sseraphini China will save us.

yo, i'm actually worried. codex limits are genuinely insane so it's sus af .. i feel this is an intentional move for a honeymoon period until we get over the claude → codex migration and then we get rugpulled hard

lol v3 of my stock trading system learned of inverse Cramer on its own

So, @rspack_dev 2.0 was launched a few weeks ago and Meteor is on the official ecosystem list 🎉 The big win for large Meteor apps: persistent cache now drops memory usage by 20%+, and SWC minimizer cache hits make builds ~50% faster. Less RAM, faster builds. We'll take it 🤝

Works quite well

You can just make gpt pro return a git patch. Download it and apply it

I saw a lot of people saying Module Federation doesn't work with Angular. Felt like there was a lack of examples out there. So here's two: one with @vite_js 8 and another with @rspack_dev Rsbuild.

@yagiznizipli @X I’ll help you guys move off webpack

We're slowly getting there with the performance of @X web (lower means better)

@oliver_bauer @nstlopez No discord. Just DMs in discord

@ScriptedAlchemy @nstlopez Can I also join the discord ? Currently building my second trading after a first on trend / news follower strategy

Going full quant and training financial model. Stealing some ideas from TikToks recommendation engine around realtime adaptive training of the model.

@Tylopilus @nstlopez No just DMs

@ScriptedAlchemy @nstlopez Wait you have your own discord channel?

@jeffscottward Yeah I use them a lot. What would be nice is if this were “stacked” worktrees since usually work has dependencies on other agents. So restacking and rebasing a worktree stack would be quite useful

Is there anyone out there who uses git worktrees really aggressively or agent orchestrators? Like creating multiple branches within a work tree off of a base repo? ———— We are potentially cooking up something really crazy in Maestro and wondering if more than one branch per work tree makes any sense realistically? Typically it’s one feature branch per Work tree right? @ScriptedAlchemy @theo @kenwheeler @elonmusk @jbrukh @Jason Please Retweet! #git #worktree #ai #ui #agent #orchestration #claude #codex

@nstlopez Check discord message. I sent you the “master plan” that contains the system end to end. From project creation to deployment

@ScriptedAlchemy I need to up my game and get into training myself

@derekdamko Can likely get another 1tb vram if necessary.

@ScriptedAlchemy You do not have enough VRAM for any of the large open weight models. Example GLM 5 is 1.5 TB on disk which will 1.7 loaded into ram. Then you will need a large context window so you are probably looking at close to 2TB of ram needed for fine tuning.

I have a B200 cluster. Looking to fine tune a model. What’s the best OSS model that’s near GPT 5.4 or opus 4.7

@omniwired Cost. I wanna fine tune a new tokenizer encoder/decoder. Only have enough burn for 1 or 2 runs.

@ScriptedAlchemy Is it coding? Then GLM 5.1 But why only try one, do Kimi and deepseek pro next. Fine-tune to do what

@HOARK_ Turboquant for inference tho, not training. But yes.

@ScriptedAlchemy i would do 5.1 with turbo quant

@ptremblay Want a big 16fp model. Like GLM or kimi

@ScriptedAlchemy what about a 1.58-bit distill of Qwen 3.6 27B? (BitNet Distillation) arxiv.org/abs/2510.13998

@julianharris 1tb of ram, can get more if needed tho.

@ScriptedAlchemy Not sure how much RAM you have or whether you are ok with open weights but Deepseekv4 is killer.