this continues to reinforce my belief that current frontier models are completely terrible at medical imaging analysis DO NOT TRUST LLM's interpretation of your medical images!!

Ppl in the comments piling up on BURKOV but there are dozens of fine-tune services, but better bc * no waitlist * more base models * more modalities And they all think they can make $$ from this low frequency tasks

When we published in the Cerebral Valley newsletter in January, they said it was among the highest engagement editions with over 54% open-rate Have another edition this week with a more complete vision of what we’re building at Remyx AI So go subscribe!

The AG2 team liked our presentation on GitRank and has shared our work in their recent ones We’re planning another presentation soon and a collaborative blog post with the team If you wanna see the deep dive on GitRank, go to YT: m.youtube.com/watch?v=N_FNfZ…

The way they’ve handled citations leads the reader to believe it’s the first or even best work applying agents to build environments for research code But we’ve documented this all over social media and have launched a service before this was published x.com/smellslikeml/s…

🚫 Another paper showing current multimodal models are not yet reliable for pointing out exact disease spots on medical images. On chest X-rays best model (GPT 5) reached only 49.7% average localization, radiologists were at 80.1%. They split each X-ray into an 8x8 grid and ask

Join us at Experiment 2025 We're hosting at Frontier Tower on Oct 30th Excited to meet world-class experts in experimentation, the builders of the agent frameworks we use, leading voices covering what's happening in tech, and you too! Sign up: luma.com/experiment-2025