0xserenade

Localllm and agentic coding. Hardware: 2x GB10, 1x M3 Ultra 256 gb

[0xserenade @pgserenade]

@mattshumer_ @benatcortexai "Rare"? Most if not all enterprise use cases are this. Government and financial use cases are massive. You would't know if you aren't in the frontline and all you do is write checks with LPs money

[0xserenade @pgserenade]

@sandboxws @ThePrimeagen Forgot to take your meds before logging in huh?

[0xserenade @pgserenade]

@antigravity It took google PM half a year to add plan mode, we have all moved on

[Toven @pingToven]

ragebaiting theo is a great virality strategy

[0xserenade @pgserenade]

@hunvreus Actually would be nice to get a deep dive on what those projects he made actually does and judge the code and features. Seems like thats up for grabs and at least sound entertaining- get astra to review and roast it

[0xserenade @pgserenade]

@Beamsters1 @LittleNonsuch Got a 256gb box, how much of performance here will i get?

[0xserenade @pgserenade]

Saw at least 4 sandbox product launch in the past week alone. You are just a service for a shared CPU rental.. been around since the 2010s. a lowest tier EC2 instance dressed up as a "Agentic Native Runtime" trench coat

Bro you are just someone else's computer

[0xserenade @pgserenade]

@bebraw @mitsuhiko Current 27b is already significantly better than Opus 4

[0xserenade @pgserenade]

@0xEricBrown +1 all future opensource is Pi shaped. Built for plugins, hooks, and forks to bring your own functionality

[0xserenade @pgserenade]

@MiaAI_lab People who owns 4 sparks, is there a marginal return on earning >2 sparks? In term of usage and ability to run high concurrent tasks/big models

[0xserenade @pgserenade]

Based, but on the flipside: demand from frontier human will likely have 95th percentile of token budget and usage ie. Auto research, benchmarks, trading... their roi from frontier token is much higher so they will spend 1000x more than your avg. Copilot user

[🔮(𝕏ᵀ𝕏) - loracle.hl @loraclexyz]

People are going to realize it takes a frontier human to extract productivity out of a frontier model, and there is simply not enough frontier human to compensate for the capex relying on revenue from training frontier models (yes if you remove training compute demand the whole thing is unprofitable already, and that’s before token price depreciation). Put it simply, demand for intelligence isn’t infinite and is in the shape of a bell curve, and cheaper models are slowly eating their way into the bell curve, leaving frontier training cost rely on a smaller and smaller tail of intelligence demand. In the end you are left with a billion dollar training run which only use is to solve Navier Stokes, because all the other tasks that are useful for the current economy can be run on cheaper models

[0xserenade @pgserenade]

@Beamsters1 @ivanfioravanti As an M3 ultra user, I will wait patiently for this

[0xserenade @pgserenade]

@sudoingX Imagine having to pay rent for your own brain for a $20

[0xserenade @pgserenade]

@MiaAI_lab He is just talking his books (just look at his bio), his portfolio and his LPs rely on it. His entire work is to appear bullish cloud models to keep the money flowing

[0xserenade @pgserenade]

Programming as a profession is dead, but its a better time than ever to program as a hobby.

Leet code is literally a video game

[0xserenade @pgserenade]

@Savlambda @effectfully If functionality and invariants are formally verifiable, it would be simpler for machines to generate raw machine code from scratch, every time as part of a build process than skirting meat space SDLC that we have to do around extending and maintaining source code

[0xserenade @pgserenade]

@GPTWare @loktar00 @MikeBradleyAI He did, he got free engagement

[0xserenade @pgserenade]

@magicblock @solana @SolanaEvents @SolanaFndn @SuperteamTH >14 teams spent a week in Bangkok

Imagine the smell

[0xserenade @pgserenade]

@Beamsters1 Gave you a vote 🍣