Tomás Marques

Creative Technologist

[Tomás Marques @tomasmcm]

Hy3 is goooood, it has a curated taste that I haven’t seen in models this size (300B). It does not try to be too smart, but still considers a lot of scenarios (looking at the thinking logs). You can try it for free with OpenCode.

https://t.co/RGsPfkFQPs

#ai_guild

[Tomás Marques @tomasmcm]

You might have seen the Anthropic post about J-Space, and now we have the same interpretability for open source llms 💥

https://t.co/Ubssx2hfu5

#ai_guild

[Tomás Marques @tomasmcm]

Manage all your TUI agents in the same app

https://t.co/llCT7n6rTR

#ai_guild

[Tomás Marques @tomasmcm]

@sflorimm there are still plenty of "frontier" models with high hallucination rates https://t.co/tFK4hL28Hk

[Tomás Marques @tomasmcm]

Is this what we are doing after harness engineering now?

https://t.co/dnRPUO3cXP

#ai_guild

[Tomás Marques @tomasmcm]

Mac OS 27 will ship with a 20B sparse model “AFM 3 Core Advanced”, not MoE. For each prompt, a router picks a set of experts and only those (1-4B) are loaded into ram (instead of per token). They are calling it “Instruction-Following Pruning”.

https://t.co/2co6x7tYZB

[Tomás Marques @tomasmcm]

Composer 2.5 inside Grok Build is faaaaaaaaast! Might be my new default

https://t.co/IE4Y6VhK6i

[Tomás Marques @tomasmcm]

Small MoE models have the potential to be exceptional in edge deployments, especially when combined with dynamic loading inference like FlashMoE. Glad someone is still making those

https://t.co/9ZWmCYWz8v

[Tomás Marques @tomasmcm]

PrismML’s Bonsai Studio iOS app allows generating images fully local using their new Bonsai Image diffusion model using FLUX.2 Klein 4B architecture.

https://t.co/HEpq3OL4Ns

[Tomás Marques @tomasmcm]

Grok Build... yeah I think that's the answer 😅 everything alright Grok?

[Tomás Marques @tomasmcm]

The new Cohere Command A+ might not beat the SOTA models but it sure has a lot less hallucinations 👏 nice!

https://t.co/hce3w32pDc

#ai_guild

[Tomás Marques @tomasmcm]

This is an interesting shift in the current Dev world. With AI we no longer need an hackathon to jumpstart new projects, but it’s a great environment to find a team, evolve / pivot a product or contribute to open source.

https://t.co/yk6DpRHfcT

#ai_guild

[Kartik Talwar @kartiktalwar]

https://t.co/BCivlyHXiR

[Tomás Marques @tomasmcm]

Great short explainer on how Diffusion Models work and are trained to generate images 🤓

https://t.co/aDmKO9A1w7

#ai_guild

[Tomás Marques @tomasmcm]

Hermes Agent now has a built-in kanban board to manage agent workers ❤️

https://t.co/9FJyAII8kF

#ai_guild

[Tomás Marques @tomasmcm]

@luismbat @rtwlz Tried a couple but always says no matches found. Since you are collecting all this data you might as well show how many tickets per movie are already bought (the less than 3 rule might be too restrictive)

[Tomás Marques @tomasmcm]

Mistral Coding Agents can now run in the cloud, their 128B dense model is open weight at Sonnet level, and the Web UI can research and create / edit docs. Nice to see Europe catching up 💪

https://t.co/HPa1bhVqFS

#ai_guild

[Tomás Marques @tomasmcm]

Google Docs for collaborating with AI agents ❤️

_When it’s easier to build a new platform than integrating with existing ones_

https://t.co/BmvKNiORWK

#ai_guild

[Tomás Marques @tomasmcm]

I remember when DeepSeek R1 was launched, the thinking was a long list of “Wait… Hmm… Okay… Also…” and now Qwen was able to teach a model HOW to think 💥

[Tomás Marques @tomasmcm]

The new Qwen3.6-27B is fascinating because of how it thinks:

The model follows a Plan > Execute > Review > Refine process for EVERY prompt.

It can think through and catch its mistakes inside a single inference run, it’s Claude Code reasoning built-in!

#ai_guild

[Tomás Marques @tomasmcm]

OpenRouter has a new labs function that calls multiple models at once and then merges the results, pretty cool

https://t.co/xWaY9kO0bG

#ai_guild