TM retweeted
One of the most interesting studies on regret found that our biggest regrets aren't usually failing to meet our obligations.
They're failing to become the person we thought we could have been.
There are people who leave a mark on the world, and then there are people who change it entirely. Steve changed the world and so many lives in the process. He certainly changed my life forever. His spirit lives on in everything he created and everyone he inspired.
TM retweeted
엄청난 결과물.
Claude가 18시간을 들여 만든 영상, 제목은 ‘Macrohard: Windows XP’.
Seedance와 Midjourney 영상을 에셋으로 활용, Windows XP 데스크톱에서 뮤직비디오를 녹화하는 것처럼 상상하라고 지시했습니다
영상 모델 크레딧 약 190달러와 Claude Code 구독 200달러의 절반 정도.
TM retweeted
Strongly agree with the picks and shovels approach and that CROs are the most obvious winners in the short term. We will see the number of compounds being generated by AI explode and the pressure now shift towards physical testing.
The focus of AI Bio is squarely aimed at coming up with drugs right now but I think in the long term the more valuable application is in drug testing & medical diagnostics. Any new technique for testing drugs that increases the likelihood of passing a clinical trial should be more valuable than just the drug themselves, similar logic also applies for identifying targets.
In the future once we know the target creating a drug candidate will probably be as trivial as prompting Codex which means the edge shift towards testing & verifying them more accurately and quicker!
AI Drug Discovery Is Becoming a Bottleneck Trade
I get excited when an industry starts going through a real regime change. AI drug discovery (“AIDD”) increasingly looks like one of those moments.
Two things are happening:
1/ Upstream: the frontier AI labs are piling in.
2/ Downstream: The supply chain is clearly moving.
Supply-chain checks suggest the upstream picks-and-shovels of discovery and early preclinical R&D are starting to feel the increase in experimental volume.
DNA → protein → assays → sequencing → automation → preclinical testing
Names across that stack include $TWST , @GenScript , $ILMN , $TXG , lab-automation vendors and CROs.
$TWST expects triple-digit percentage growth in AI-enabled drug-discovery orders in FY26, and another year of triple-digit order growth in FY27.
@GenScript's AIDD business doubled YoY in 1H26. Its current platform advertises industrial-scale validation of 4,000+ designs/day, with integrated sequence-to-data workflows. Our channel checks suggest the ramp is moving even faster: roughly 8,000 designs/day currently, with a path toward ~16,000/day by YE26.
---
Why does AI drive more wet-lab demand?
1/ AI makes hypothesis generation almost free → way more shots on goal.
The bottleneck is moving from expert-driven design to biological validation.
2/ AI models need continuous experimental feedback — and both good and bad data are useful.
Traditionally, only the highest-conviction A+ candidates might get pushed into expensive validation. With AI, even the B/C candidates can be valuable because failed experiments generate training data.
@GenScript has said its sequence-to-binding workflow can return data in 4–7 days, and that faster cycle times matter because AI models depend on continuous experimental feedback.
@Anthropic is a clean example. @claudeai designed 1,320 protein binders. @adaptyvbio converted those digital sequences into DNA, expressed the proteins and tested binding. Only 354 actually bound. And the 966 failures are not wasted. They are useful negative labels: what does not express, what does not bind, what has poor affinity. Those results help train the next model iteration.
3/ Wet labs are no longer just making drugs. They are making training data.
$TWST / @GenScript increasingly look like biological data foundries.
$TWST explicitly talks about generating model-ready data from AI-designed sequences. In some workflows, the customer may care less about receiving the physical protein than about getting structured experimental results back into the model.
Traditional drug discovery asks: “Does candidate X work?”
AI drug discovery also asks: “What can this experiment teach the model?”
---
TAM of AIDD
If AI is simply a better R&D tool, the relevant spending pool is the $300–400B of annual global pharma R&D. If AI meaningfully increases the number of viable drug programs, the opportunity is larger because it expands downstream demand for DNA synthesis, protein production, assays, and preclinical work.
Near term, we can also size demand from AI-company spending. If Anthropic reaches $80B of ARR in 2026 and spends just 1% on AIDD, that alone would imply ~$800M of annual investment.
---
Trade setup
This is a trade that could have long legs. It’s hard to really stop working until PhaseI/II results (2028+)
It smells a lot like the bottleneck trade we just saw in semis: GPUs → HBM → networking → power/cooling. In biology, It basically follows the drug discovery process downstream: AI models → designs → DNA/protein → assays → preclinical capacity.
After the upstream picks-and-shovels, animal testing could become the next bottleneck. Monkey prices are already near prior highs and CRO capacity is tight. AIDD pushing more candidates into preclinical development would only add demand.
?? But clinical trials are still the bottleneck?
This is the biggest pushback I keep coming back to. No matter how fast discovery becomes, drugs still need to go through preclinical → Phase I → Phase II → Phase III → approval. You still need patients, time and capital.
But that doesn’t mean the bottleneck trade won’t work. More viable candidates — especially with higher success rates — still means more demand throughout the development process.
And who knows: clinical trials themselves may eventually be optimized by AI.
?? What breaks the trade?
Near term, the picks-and-shovels trade breaks if experimental budgets stop growing, AI-generated designs don’t translate into useful wet-lab hits, or capacity catches up too quickly.
Longer term, the thesis breaks if AI drugs look great in discovery / Phase I but fail at normal rates in Phase II/III. That is why Phase II matters so much.
?? Milestones
Late 2026–2027: first Isomorphic-designed drugs enter human trials; more AI-native programs move into IND-enabling work / tox.
2028–2030: clinical trial results start telling us whether AI-designed drugs actually perform better than conventional drugs.
Calling all the "bottleneck bros". :) @jukan05 @zephyr_z9 @aleabitoreddit @ParadisLabs
+++
More comprehensive analysis: robonomics.substack.com/p/ai…
TM retweeted
AI will solve aging, aging reversal is coming
AI designed drug shows signs of reversing biological age
Insilico Medicine’s AI discovered and AI designed drug rentosertib, originally developed for idiopathic pulmonary fibrosis, produced a striking secondary result in humans: all 6 proteomic aging clocks measured patients as biologically younger after treatment.
The analysis involved 42 patients in a Phase IIa trial. The strongest regimen showed roughly 3-4 years of predicted biological age reversal after just four weeks, with one aging clock indicating as much as ~6 years.
AI identified the biological target TNIK, and generative AI then designed the drug molecule itself.
The drug has now advanced toward Phase III development for lung fibrosis.
TM retweeted
reading feynman lecturess as feynman intended: with accompanying ai-made demo sections for each chapter
rapid acquisition of knowledge is possible thanks to superintelligence
TM retweeted
Most exercise and aging studies can't answer a basic question: is muscle deterioration from aging itself, or just from decades of moving less? A new Nature Aging study solved this by recruiting older adults who moved as much as people in their twenties.
The researchers from Amsterdam UMC and Maastricht University recruited four distinct groups: young adults in their twenties, older adults whose daily step counts and high-intensity activity matched the young group, older adults who had trained consistently for years (three structured hour-long sessions per week for over a year), and older adults with early physical impairment.
They took muscle biopsies before and after a one-hour cycling session, then measured over 24,000 gene transcripts, 135 metabolites, and 1,383 lipid species.
By matching activity levels between young and older groups, any molecular differences couldn't be blamed on the older adults simply moving less. This isolated aging from inactivity for the first time at this molecular depth.
Key findings:
• The defining molecular signature of muscle aging is an energy crisis. Comparing young adults to activity-matched older adults, 1,106 genes were downregulated with age. These genes build the mitochondrial machinery that produces cellular energy: ATP synthase, cytochrome c oxidase, and NADH dehydrogenase subunits. Mitochondria are the power plants of cells, converting nutrients into ATP, the energy currency cells use to function. When these genes decline, cells lose their ability to generate energy efficiently.
• NAD+ levels declined and triglycerides accumulated inside aging muscle. NAD+ is a molecule required for energy production and cellular repair. Lower NAD+ means less capacity to convert fuel into usable energy. Triglycerides are stored fats, their accumulation inside muscle indicates unburned fuel piling up as the tissue loses its ability to process it.
• More than half the molecular signature of muscle aging was absent in trained older adults. Specifically, 57.1% of age-related gene downregulation and 55.9% of upregulation were missing in the trained group. Their muscle resembled young adults far more than their chronological age would predict.
• The changes training preserved were specifically the energy metabolism ones. Genes like NDUFS1 and COX5A, which were depleted in normally active and impaired older adults, sat at youthful levels in the trained group across all five mitochondrial respiratory complexes. The single most prominent feature of muscle aging turned out to be the single most preventable.
• Being generally active was not enough. Structured training was the difference. The normally active older adults walked as much as young adults, and their energy metabolism genes declined anyway. What preserved the youthful molecular profile was structured, sustained training. Filling a step counter and being genuinely trained are not equivalent at the molecular level.
• Roughly half of muscle aging persisted regardless of training. Changes in genes controlling synaptic transmission (how nerves communicate with muscle) and WNT signaling (a pathway regulating tissue maintenance and stem cell function) appeared in all older adults, trained or not. This unavoidable half is where drugs will have to work.
• The fittest muscle mounted the largest inflammatory response to exercise. All groups activated stress and immune genes after exercise, including IL6, IL1B, and TNF. But the magnitude scaled with fitness. Trained older adults most closely resembled young adults in their response, followed by normally active, with impaired older adults showing the most blunted response. The stress response to exercise appears to be the mechanism of adaptation, not damage to be minimized.
This raises a concern about anti-inflammatory longevity strategies. If the inflammatory stress response is how exercise produces its benefits, chronically suppressing inflammation may blunt the adaptation that exercise depends on. It doesn't mean inflammation is beneficial in general, but the timing and context matter.
A separate discovery: the proteasome appears to regulate NAD+. The proteasome is the cellular machinery that breaks down damaged proteins. When researchers inhibited it, NAD+ levels rose in both muscle and liver cells to a degree comparable to NAD+ precursor supplements. This opens a new route to understanding NAD+ decline that operates through protein turnover rather than just supplying more raw material.
The study draws a clear line between what lifestyle can address and what will require therapeutics. The energy metabolism decline, mitochondrial deterioration, and NAD+ depletion that define muscle aging are largely preventable through structured training. The synaptic and signaling changes that persist in all older adults represent the unavoidable half where drugs will need to work.
The decisions made about structured training in midlife determine which molecular trajectory muscle follows in later decades. Half of muscle aging is optional. The other half isn't. Knowing which changes belong to each category is knowing where behavior ends and biology takes over.
TM retweeted
Polyethylene contains approximately 46 megajoules of chemical energy per kilogram (about ten times the energy content of TNT by mass) stored stably in its carbon-hydrogen bonds rather than as explosive potential.
The Oak Ridge molten-salt process announced this week is essentially an engineered way to release that energy as liquid fuel rather than letting it sit in a landfill.
TM retweeted
Our lab’s v2 in vivo age-restoration technology is fundamentally different from the v1 technology now in human trials. We’ll reveal it publicly at @ARDD_Meeting: Chris Petty, Thurs 1:30 p.m. ET, and me, Sat 6 p.m. ET
Replying to @davidasinclair
I think there is misunderstanding that OSK or OSKM accomplish all the effects through resetting epigenetic age clock only. That may be a small part of what they are capable. OSK/OSKM opens chromatin’s 3D structure and activate all sorts of genes that are not accessible
TM retweeted
Claude 페이블 5.1도
Claude Opus 5.5도
그리고 GPT-6 Astra도
effort(추론)를 Max로 올린다고 항상 좋아지지는 않습니다.
그리고 이걸 이해하면 "토큰"도 제대로 아낄 수 있습니다.
TM retweeted
Scientists uncover thousands of previously missed proteins, many of which appear to be relevant to human health — including promising new targets for cancer immunotherapies nature.com/articles/s41587-0…
rdcu.be/kUYzGHZ13eea
TM retweeted
OpenScience is now the #1 scientific agent.
Today it's out of beta and live on Product Hunt, with:
• A new IDE: a faster, fully redesigned research workspace
• OpenScience Ace: 30+ hand-picked models, including GPT-6 Astra, Claude Opus 5.5, Grok 4.7, Kimi K3, GLM-5.3, and DeepSeek V4.1 Flash, in one pay-as-you-go wallet
• Autoresearch: give it a metric and it hill-climbs through experiments on its own
• One-click OAuth to bring your ChatGPT or Codex subscription
• 300+ research skills, 50+ scientific tools and databases, and NVIDIA BioNeMo built in
Already in use at 30+ universities and research labs. Free and fully open source.
애플코리아여, 언제까지 사용자가 목말라서 한글 환경을 뜯어고쳐야겠습니까?
어쨌든 한영 전환을 빠릿빠릿하게 해준다는 한영 전환 유틸리티입니다. github.com/codingnoye/gksdud
Artificial-intelligence systems can generate hypotheses, design experiments and analyse data — but humans still need to decide what makes sense
go.nature.com/4rPJHYl
TM retweeted
We’ve set up a molecular biology lab at Anthropic and we’re announcing our first discovery! Claude discovered a new CRISPR-like enzyme.
950 agents spent 21 hours searching through a database of DNA sequences until one of the agents found something striking: “[The DNA next to the RT] is spectacular: I can see by eye a tandem repeat array … that's a CRISPR-like … repeat array?!”.
After analysis and testing in our lab, we found that the sequence is a previously uncharacterized enzyme system. We don’t know what it does yet, but it has features reminiscent of CRISPR.
Our lab looks like a typical molecular biology lab. Our research only involves the lower-levels of biosafety risk level; we don’t handle pathogens that can infect humans, and all the lab work is performed by human scientists. We’re sharing these early findings with the community to show how Claude can be used to accelerate fundamental research in biology.
Claude has discovered a previously unknown enzyme system hidden in the DNA of bacteriophages. Beside the enzyme’s gene sits a long array of repeating DNA—a structure that looks somewhat similar to CRISPR.
We don’t yet understand what this system does, but only a handful of known systems share its features, and all of them are able to cut, copy, and paste DNA. Historically, the discovery of such programmable systems has helped revolutionize medicine. CRISPR, for instance, is now the foundation of genetic medicines. But it will take much more work to learn what this system does, and whether it can be put to similar use.
Read more: anthropic.com/news/claude-di…