BarbarianJan

@memelord197 @ahitposter building what?

@ahitposter Lol he’s pushing like 5000 commits on good days, it’s impossible to keep track of all the updates and bug fixes

> “so i see here you are spending $1.3m per month on ai credits” “yes that’s correct” > “what exactly have you built this month that warrants this kind of spending” “we built this nice ui that shows you how much we are spending on ai credits”

@tit0jacks0n @GaryMarcus interesting - how you manage that?

@GaryMarcus oh another i thing -- i have it catch / sort / summarize all of the things i encounter scrolling so they end up in the right place in obsidian. so youtube videos, tweets, reddit convos, blog posts, articles, etc. a memory-keeper that helps me feel less overwhelmed.

Who out there is using Hermes, and for what?

@BigSteelShips @JDabknee he can talk with GPT for all i care

@JDabknee This guy's like AI homelander, he just needs someone to talk to LMFAO 😭😭😭

@sudoingX carnice v2? :)

hear this. carnice v2 on hermes agent, rog scar 5090 mobile at 99% gpu, 59°C, sucking air from every direction. this is what local ai sounds like at full load.

@bysahilhooda @adamghowiba why do you say it starts getting unfair?

Yeah this is basically the blueprint now. Not one “smart” agent doing everything, but a system: One agent thinks Others do specific jobs Another checks the thinking Human signs off It’s less like using AI and more like building a small company inside your workflow. The interesting part is this isn’t just for big firms anymore. Anyone who understands this properly can build their own little decision engine. That’s where it starts getting unfair.

JP Morgan's investment research team just shared exactly how they built their multi-agent system "Ask David", and it's the same architecture pattern showing up everywhere: - supervisor agent orchestrates - specialized subagents handle retrieval, structured data, analytics - LLM-as-judge reflection node before the answer ships - human-in-the-loop for the last accuracy gap worth watching for anyone building:

@sokol_cc go local or perish

Is it just me, when I talk to artificial intelligence about my ideas and technical solutions, I don't say everything because I think that someone on the other end might get a notification about an "interesting idea"?

@CKeruac @raiden0010 imo czas na lokalne modele - nie dawać się dymać frontier labsom

@raiden0010 fakt ostatnio codex tez słabo działał - serwery przeciążone :)

AI właśnie wjechało w ścianę. Pure scaling umiera. Coraz większe modele dają coraz mniejsze efekty. Euforia 2023-2025 się skończyła.

@RetroChainer what is the iphone setup?

> lawyers see a patent close the tab > Claude sees a patent calculates margin > Mac Mini. Claude Code. one evening. > $1.80 to make. $11.99 to sell. > 44% margin. zero R&D budget. > vibes don't compound. patents do.

@vtchakarova spooky

Link gfkv.org/wp-content/upl…

GLOBALER SYSTEMBRUCH: Die vierte systemische Krise und ihre Architektur (in Deutsch). Link in den Kommentaren!

@LottoLabs @glitch_ oh nice @grok which website are doing the above?

@glitch_ Yeah I’d probably use a burner laptop, then remote into my node, I guess it depends what you’re doing. If you need local that’s the way, if you’re testing you could just rent a 3090 depending on what you’re doing, they’re like .40$/hr

@SelectedCinema Matrix vibe...

𝐁𝐥𝐚𝐜𝐤 𝐌𝐚𝐬𝐤. 𝟷𝟿𝟿𝟼 / 𝙳𝚊𝚗𝚒𝚎𝚕 𝙻𝚎𝚎

@AndrewCurran_ FTX wibes,

OpenAl is offering private-equity firms a guaranteed minimum return of 17.5%, as well as early access to models not yet in public release.

@sudoingX can you outline what is your full setup? Not only hardware but software as well along with your hermes/qwen setups? :pray:

also just opened my DMs to verified users. was locked before and didn't realize. if you tried to reach me and couldn't, try again.

let me be very clear. these little models do more than i show on X. if i told you what i squeeze out of 9 billion parameters on 12 gigs of VRAM daily you would not believe me. things i can't even post because the results would look fake. i use one prompt and one game as my standard benchmark because it forces every model through the same test. if a model can build this autonomously on one GPU it can build anything you throw at it. the game is how i show you the scale. what's running behind it is the real story and that's coming next.

@frostyflak3s @AzFlin can you expand please ?

@AzFlin 99% people dont even need agents or openclaw

You 👏 Do 👏 Not 👏 Need 👏 Multi-Agent 👏 Orchestration 👏 Systems 👏

@mert /gi

some 2nd order predictions from this: - the rich gulf states will almost certainly start incentivizing defense startups. they have trillion dollar investment arms. - specifically, since they are tiny nations, they need autonomous weapons. they can't field large land armies. - but anthropic/openai doesn't allow for autonomous weapon military use, so it is likely the gcc start using chinese AI models - but that is obviously unfavorable for america, so then they'll have to either i) make a frontier lab play ball, ii) start their own effort via help from palantir/anduril - it seems inevitable to me that this conflict directly accelerates the militarization of AI, which in turn implies gigantic state-sponsored spending into the AI supply-chain - however, there's a growing populist movement in the west around being anti AI/datacenters due to potential job loss for the middle class - I think the above tension continues to accelerate until the next USA election. and that election will decide the fate of the West

@cgtwts I've been hearing that this model was made specifically to do good on benchmarks but overall it is meh - can anyone confirm/deny these accusations ?:)

MiniMax just dropped M2.5: - 80.2% SWE-Bench (beats Opus 4.6) - 37% faster than M2.1 - $1/hr at 100 tokens/sec - cheaper than Opus 4.6 - it’s open source.

@Dirteii @OwariETH 2025 retard

@OwariETH So this would print the cycles ATL on October 5th

simulation

@lukaszb99 @WarNewsPL1 @NATO @SecRubio @grok prawda to?

@WarNewsPL1 @NATO @SecRubio Znaczy się, senator Duckworth jest senatorem z okręgu w którym albo mieszka sporo Amerykanów polskiego pochodzenia albo w tym roku będą w jej okręgu wybory do Senatu.

🇺🇸🇵🇱🇦🇫🇮🇶🇪🇸🇬🇧 Senator Duckworth: Kiedy służyłam w Iraku, były tam z nami wojska polskie, brytyjskie, hiszpańskie… Polscy żołnierze zginęli w Afganistanie. Nasi sojusznicy z @NATO stali razem z nami. @SecRubio: Prezydent chce po prostu, aby nasi sojusznicy mieli większe możliwości. Senator Duckworth: Nie to powiedział. Powiedział, że zostali trochę z tyłu i nie stanęli w naszej obronie. Rubio: Nie wszystkie państwa @NATO brały udział w tych misjach. Źródło: @Acyn

@Miraqulee @VladTheInflator @grok is this movie a nazi propaganda?

@VladTheInflator Because they lied about everything with Germany. Watch Europa the last battle for the truth.

Why did they teach students that the Germans burned books on science when really it was pornography and books about transgenderism? That sexual-science and gender-identity research was literally the central target.

@Brock_Samson254 @retro_anime Have you seen followup scenes mate?

@retro_anime Honestly this scene made me worry more about what Humanity will do to AI, than what AI will do to Humanity

The Animatrix (2003)