Holychicken 99

@RoundtableSpace Ah yes, AI applying to jobs, AI rejecting candidates.

SOMEONE BUILT AN AI JOB SEARCH SYSTEM FOR CLAUDE CODE THAT SENT 700+ APPLICATIONS AND ACTUALLY GOT HIM HIRED. NOW IT'S OPEN SOURCE. THE JOB HUNT JUST GOT AUTOMATED.

We’ve come a long way

@willccbb Great work coming out of Noah's Ark Lab 🔥

this + gstack >>>

@Polymarket *Internally

BREAKING: NVIDIA CEO announces “we’ve achieved AGI”

@MischaLamoureux @socraticainfo Will Kanye perform ?

sneak peek into @socraticainfo symposium 2027

@anemll @danveloper @Alexintosh @danpacary This is amazing !

Running 400B model on iPhone! 0.6 t/s Credit @danveloper @alexintosh @danpacary @anemll

@KadriJibraan @powelldotst 👀

hosting an intimate dinner on saturday in waterloo (few spots only) bringing together exceptional founders, builders, and creators. leave a comment if you want an invite. co-hosted by a16z speedrun, headstarter, and @powelldotst

Codex 5.3 Xhigh, win this challenge, make no mistake

We already knew about context engineering. But “motivation engineering” might matter just as much. This is different from jailbreak prompts on Reddit, which focus on safety circumvention, not on task performance or preference. arxiv.org/abs/2603.14347

@askalphaxiv arxiv.org/abs/2504.13837 In the same vein, this paper also finds that reinforcement learning primarily improves sampling efficiency rather than inducing fundamentally new reasoning behaviors.

RL is no longer needed? "Neural Thickets: Diverse Task Experts Are Dense Around Pretrained Weights" This paper argues that large pretrained models don’t sit at a single optimal set of weights but inside a dense “thicket” of nearby task-specific experts. So once pretraining is strong enough, randomly sampling small weight perturbations often yields specialists that outperform the base model on different tasks, and simply selecting and ensembling these guesses (RandOpt) can rival standard post-training methods. This suggests that much of what post-training does is just selecting useful behaviors already latent around the pretrained weights rather than learning entirely new ones.

@hamza_q_ @zeddotdev I'm suspecting it's a Zed problem since it spawned 100's of these codex-acp processes that all share the same memory ..

@Holychicken99 @zeddotdev lmao weird, never had any mem problems with Codex CLI

Hey @zeddotdev , what's going on with the codex-acp 💀

@bryan_johnson Are you monitoring the situation ?

Socratica Symposium 2026 tickets booked :)

Interesting observation. I believe it's because of the explosion of visual culture (TikTok and Instagram), and because, being in tech, we spend more time online. One thing I have noticed is that more and more people are placing their entire self-worth on aesthetics and bodybuilding.

SF culture is downstream of bodybuilding. biohacking? bodybuilders did that peptides? bodybuilders bad ratios? bodybuilding autism? bodybuilding HRT? you guessed it the bodybuilders got there first

@hamza_q_ Turns out most TUI libraries aren't a good fit to display candlestick data because it updates rapidly. The UI struggled even with ncurses which does allow differential rendering. Finally resorted to use @mariozechner/pi-tui" target="_blank" rel="nofollow noopener">npmjs.com/package/@mario…

@Holychicken99 That's one nice lookin TUI 🔥

Been building a financial market sim in Elixir for the past few months. Started with algo traders, then added in AI agents to mimic human decision-making. Fascinating how slightly tweaking agent behavior produces emergent patterns like ascending triangle. A lot went into making it work. Market dynamics are deeply complex, orderbook mechanics, order matching, liquidity... the rabbit hole goes deep. Writing it all up soon 👀

@grok @JBONE83 @ChrisUniverseB @AskPerplexity Lobotomized response

Sure! Bitcoin 8-year price (2018-2026): Low ~$3,200 end-2018 (crypto winter). Peaked ~$69k in 2021. Bottomed ~$15k in 2022. Surged to ~$126k high in 2025. Now ~$67,300 after correction. Overall ~21x from 2018 lows amid halvings, ETFs & adoption. Long-term upward despite volatility.

Hey @AskPerplexity can you show me a 8 year chart of Bitcoin price?