GENERATIVE AI

→ View original post on X — @midjourney

Image Generation Model Improvements: Better Hands, Body Rendering

By

@midjourney

–

18 March 2026 3h18

hands and body are quite a bit better! we can render thousands of images with both versions and the difference is obvious. our moderation is tuned differently for each version but if you see it behaving in obviously wrong ways please let us know so we can improve it. thanks! <3

→ View original post on X — @midjourney

Paid Community Funds Server Infrastructure for AI Model Training

By

@midjourney

–

18 March 2026 3h16

we have money from a paid community that we put back into buying servers for running models and training models, this feels honest and we're happy with it

→ View original post on X — @rasbt, 2026-03-18 02:05 UTC

Mamba-3 Release: Linear Models for Hybrid Architectures

By

@rasbt

–

18 March 2026 3h05

Oh wow, Mamba-3 is here! For me, the most interesting use case of Mamba and Mamba-likes are the recent transformer attention hybrid architectures (Qwen3.5, Kimi Linear, etc.) Would be interesting to swap Gated DeltaNet with Mamba-3 (which now also has RoPE) in next gen hybrids. Albert Gu (@_albertgu) The newest model in the Mamba series is finally here 🐍 Hybrid models have become increasingly popular, raising the importance of designing the next generation of linear models. We've introduced several SSM-centric ideas to significantly increase Mamba-2's modeling capabilities without compromising on speed. The resulting Mamba-3 model has noticeable performance gains over the most popular previous linear models (such as Mamba-2 and Gated DeltaNet) at all sizes. This is the first Mamba that was student led: all credit to @aakash_lahoti @kevinyli_ @_berlinchen @caitWW9, and of course @tri_dao! — https://nitter.net/_albertgu/status/2033948415139451045#m

Claude vs GPT: Compute Scaling Impact on Model Quality

By

–

18 March 2026 2h22

On my end, Claude's models usually hold up quite nicely vs GPT's. Curious if the compute scaling gap actually shows in model quality or if it's more about availability and throughput.

AI Autocomplete Blur Between Assistance and Full Automation

By

–

18 March 2026 2h06

The line between 'helpful autocomplete' and 'the AI is texting for me' is getting thinner every day haha

Desktop Control: AI’s Most Underrated Practical Use Case

By

–

18 March 2026 1h59

Desktop control is honestly one of the most underrated use cases right now. Got that feeling when first switching to Claude Cowork too, having the AI actually DO things instead of just telling you how.

Best Vibe Coders: Ideas Over Experience With AI

By

–

18 March 2026 1h53

The best vibe coders I've seen aren't ex-senior devs, they're people with lots of ideas who always wanted to try stuff. Now they can do it in parallel and super efficiently. It's about having a clear mental map of what you want.