Felipe @_dev_pipe

🔴 ¡¡OPENAI ANUNCIA GPT-6!! El nuevo modelo GPT-6 Astra ya está aquí, con un salto en capacidades MUY sorprendente! Os iré desglosando y analizando todos los detalles en este hilo 👇🧵 ¡deja tu RT para apoyar!

Claude Fable 5.1 is insane. i know literally NOTHING about coding. ZERO. and i just built 3 fully functioning web apps in 30 minutes. http://localhost:3000/ http://localhost:8000/ http://localhost:5000/ check it out.

September very well may be the BIGGEST month in AI ever... We're expecting: (Sep 1–6) -Fable 5.1 (rumored tomorrow) -GPT 6 Astra (rumored Thursday) -Grok 4.7 (Sep 7–15) -DeepSeek V5 -Composer 3 -Gemini 3.8 flash (Sep 16–22) -Kimi K model (not Kimi K4) -Potentially Opus 5.1 (Sep possible but who knows) -MiniMax 2.7T open-weight model -Meta Llama 5 / next Muse “Watermelon” -Mistral frontier open MoE Nobody knows exactly when all these will land but regardless it's going to be a INSANE month. Which are you waiting for most?

Proprietary AI is getting speedrun in real time. Qwen 3.8 Flash Next (125B A6B MoE) on the Artificial Analysis Agentic Index: - Beats Kimi 3 (Max) - Sitting just 1 point behind Claude Fable 5 - Hardware: Single RTX 3090 / 4090 (24GB VRAM) + 100GB RAM Matching top tier API performance on a single consumer GPU setup is insane. Open weights are moving absurdly fast. Are you still paying top dollar for closed APIs or self hosting?

Alok @analogalok

a week ago

The VRAM barrier is officially dead. I just ran Qwen 3.8 Flash Next (MoE) 125B A6B with a 250,000 context window on a single 24GB RTX 4090. 21 tokens/sec decode. 364 t/s prefill. no mtp. no dflash. no kv cache quantization! We are running datacenter models on consumer

160 454 5K 1.0M 5K

x.com/i/article/2093…

¿Habías oído hablar de un Discovery-to-Delivery Engineer? ¿No? Normal, tuvimos que ponerle nombre al perfil que buscamos. Alguien que no necesita un stack, a quien no le asusta un problema y no para hasta verlo en producción. ¿Ese eres tú? Inicia tu proceso de hiring.

The photo above is a coding harness tier list... and I somewhat agree with it. However, Hermes Agent @NousResearch in particular is sits a tier above even the top tier. While I respect tools like omp and Claude Code, I feel like you only need this one. So I wrapped it into oh-my-hermes and optimized it with engineering coding tools and a memory system. Maintaining codebase quality is crucial, and so is sharing context with the user. If the harness and model output bad code, they will generate even worse code on top of it, causing the system to collapse over time. I let it run with clearly defined quality Evals and goal baselines, and I have also embedded reinforced software architecture thinking—like OCP, coupling, and modular interface design. Code search efficiency is boosted by ast-grep, sub-agent routing directs tasks to the right models, and specialized prompting skills leverage the unique strengths of each model (Opus, GPT, Kimi, Fable, GLM, DeepSeek, Qwen, Grok, Gemini...). It also features parallel tool calling & Eval execution, long-term memory labeling (Fact, decision, episode, representation, TTL), and built-in interview skills for memory consolidation—all packed into a Power Intelligence Plugin for hermes-agent. I'm maxing out tens of billions of tokens using strictly this. Still have a Claude subscription instead of Nous Portal? Don't worry—just have Hermes trigger Claude and aggregate the results; I've baked in all those rules. Design skills, Ops skills—there are just tons of them. So now I work exclusively with Hermes Agent. I highly recommend subscribing to Nous Portal because I've fine-tuned task routing based on what each model does best. Since it supports multiple models, you can cancel your Claude or Codex subscriptions and just use this. oh-my-hermes is constantly evolving: github.com/rlaope/oh-my-h… I've proven that you can delete all other harnesses and achieve everything with Hermes Agent alone. GOAT HERMES-AGENT

When calories became cheap, convenient, and abundant, bodies decayed. Thinking is now cheap.

¡Google regala 1 año de Gemini GRATIS a estudiantes! ✓ Plan Google AI Plus ✓ Generación de vídeo con Omni ✓ Notebooks de estudio con tus apuntes Para España, México, Argentina, Chile, Colombia, Perú, Venezuela, Uruguay... Puedes canjearlo hasta el 31 de diciembre de 2026: → gemini.google/students/

The revised version from someone who isn't paid by any company and actually tested all these models firsthand. Being an ambassador isn't about getting paid, it is just being a beta tester to give feedback. This is the conclusion after weighing everything: performance, efficiency, and price.

Theo - t3.gg @theo

2 weeks ago

Rough tier list of where I'd put every major model right now

1K 751 15K 4.0M 5K

I always recommend Hermes Agent, and this is a good chance to try it.

Nous Research @NousResearch

2 weeks ago

Ox Alpha is free via Nous Portal for a limited time. We have capacity for 1 quadrillion tokens per day. Let the tokens flow at portal.nousresearch.com

292 240 4K 1.6M 1K

Hot take: Hermes Agent is wildly overrated. By people who’ve never actually used Hermes Agent.

If you truly understand how important AI is and the impact it’s going to have, you’d probably drop everything in your life and go all-in on it. ​Most people still don’t realize what’s coming.

@CanalMutuo Esas son respuestas compartidas por los usuarios

U.S. Poet Laureate Ada Limón’s poem celebrates water—what connects the two ocean worlds @EuropaClipper travels between. Make your special connection to Europa & #SendYourName aboard our spacecraft on this 1.8 billion-mile journey to Jupiter’s moon: go.nasa.gov/MessageInABott…