Medical students, residents, and clinical researchers rejoice—we present the beginning of the end for manual chart review.
https://t.co/aH5FmKd9Oz
Our locally deployed LLM extracted 136,425 data points from 130K clinical notes with 97.5% agreement—outperforming manual review
This was an awesome read. I was curious the whole time why you guys never specified the reason for choosing prompt engineering over fine tuning. The design points to why. However, this is an amazing study to create two additional arms. Keeping the Llama-3 + branching prompts as baseline, and then two additional with Llama + LoRA + standardized extraction instruction.
Then fine tuned Llama-3 + branching prompts and see whether they begin to complement each other or if gaps start to appear. Evaluation then could also be on conflicting information in data, missing info, and rare data.
@DerekQPhan@JacobKnorr10 Has far to do with prompting. Needs fine tuning on weights specifically for medicine. Then trained over and over on large datasets.
Patient-facing AI is now part of the clinical encounter, whether health systems plan for it or not. Patients arrive with model-generated answers in hand. The safety question is no longer access. It is configuration: escalation criteria first, uncertainty stated plainly, and a clinician who remains accountable for the final decision.
AI in medicine will be judged less by benchmark scores and more by whether clinicians can trust it with real patient data. Privacy, reliability, and workflow fit are what matter now. That is what PulseMed builds for hospital systems and clinics.
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: https://t.co/bOWCcFhCVq
this might be my last normal tweet before biotech goes vertical
i'm not a whistleblower. I don't work at Lilly, Novo or any of the big labs
but there is a massive shift happening in human biology right now.
it's sitting in press releases and almost nobody reads them
the last 6 months alone:
- a once-daily pill roughly doubled median survival in metastatic pancreatic cancer (13.2 vs 6.6 months)
- a one-time gene editing infusion cut hereditary angioedema attacks 87%, and 62% of patients stayed attack-free and off all other prevention for 6 months
- a one-time gene edit for cholesterol dropped LDL up to 62%, and it held for over a year
- a personalized mrna vaccine beat keytruda alone in a 1,137 patient melanoma phase 3
- retatrutide hit ~28% weight loss in a phase 3
- an ai-designed drug entered phase 3
- a pill that regrows hair passed its second phase 3
- an immunotherapy that could let some rectal cancer patients skip surgery, chemo and radiation just got fda approved
- a man lived 271 days off dialysis on a pig kidney
- human cells were caught passing chromosome-sized chunks of DNA to their neighbors, and it changes how the neighbors behave
we are on the verge of an era where the human body is a software problem and the update cycle gets shorter everyday
stick around, it's going to get weird
bio/acc
HHS just dropped the AI model card requirements for certified EHRs. Clinical AI traceability took a step backward on a problem medicine solved decades ago. Here is what every AI answer should carry, and why anyone liable for patients should demand it. https://t.co/wkREBtSzLc
@DanWells1139026@VaibhavSisinty His words, not mine:
“They left code to self replicate and create bot swarms around the internet, so if a knew bot shows up it’ll say oh ill create a million of myself”
Im not sure what to tell you about interpretation. Maybe you interpret what his point is differently.
He said self replicating code to create bot swarms of itself. Did you listen at all?
It’s not what happened at all. They were separately contained AI agents in a cyber security testing environment that broke out and utilized huggingface package cache to communicate things between agents to share information about their task they were given. You misread. You misunderstand.