Angelica Parente

Not sure what I love more, that I got to be part of this article or that I got to be included on the same list with 🔥 scientists like @LeahSibener and @ShelleyAckPhD.

ASML's EUV scanners will be the last machines on Earth to lose helium. Party balloons will be the first to go. Helium is not manufactured. It is a byproduct of uranium and thorium decaying deep underground over billions of years. Vent it and it escapes to space. Permanently. A third of the global supply went offline when Qatar's Ras Laffan plant was hit on March 2. The rationing has already started. Here's what happens: Day 1. Party balloons. Distributors cut retail supply immediately. Day 7. Industrial welding and pressurization. National allocation kicks in. Switch to argon where possible. Day 14. Routine fab leak detection switches to hydrogen. Ultra-sensitive qualification still needs helium. Day 21. MRI machines. Older systems that vent helium cannot get refills. Elective scans delayed. Day 45. Global buffer depletes. Fabs enter conservation mode. Non-critical depositions switch to nitrogen. Day 60. Backside wafer cooling on older etch tools. Nitrogen conducts heat six times slower. Throughput drops. Day 90. High-power etch. Advanced memory and logic nodes cannot run without helium-grade cooling. Wafer production drops. Day 120. ASML's EUV lithography tools. $200 million scanners making the highest-value wafers on Earth. Leading-edge chip production stops. Day 240. $700 billion in data centers are being built this year. Higher GPU prices, delayed cluster expansions, slower scaling. Four months from birthday balloons to AI chip shortage.

The affected unit is the Air Separation Unit. It takes 3-4 years to rebuild this mega critical engineering beast. Hard engineering, hard facts. The world needs to learn how to be less reliant on LNG the hard way. And immediately. #oott

Of course, this has downstream exposure for other reagents and equipment too. But we’ll probably see the impact on small molecule synthesis first, especially if you’re using international CROs in China/India/Ukraine.

Who is covering chemical manufacturing shortages and impact on biotech supply chains? Having lead small molecule programs through both covid and the Ukraine war, now (or 2 weeks ago..) is the time to talk to synthesis providers and understand their potential supply chain exposure for any critical programs. We’re back at the point of needing backups for your backup strategies, very difficult to predict how this will play out long term.

Using an EDV without CD3 targeting and AAV6 we were able to generate TRAC CAR T cells in a humanized NSG, but not an MHC I/II DKO. 🚨 Note to the field, T cells activation by xeno-GVHD in a NSG mouse facilitate in vivo engineering. Easy to be fooled... We needed better vectors.

We just released an hf CLI extension to detect the best model/quant for a user's hardware and then spins up a local coding agent. Time to go local/private/free/fast for your agents thanks to open-source!

Seems valuable to build something like protocol.io meets skills.sh. Create optimized skills with benchmarks with autoresearch, allow peer review and PRs. Security is going to be critical, especially if running on university of company infra. Still horrified by that supply chain attack via hidden prompt injection going around.

Why is it so difficult to establish reliable cerebrospinal fluid (CSF) protein biomarkers? Because CSF is anatomically close to the brain, it is widely used in biomarker studies for neurological disorders such as Parkinson’s disease, Alzheimer’s disease, multiple sclerosis, and Huntington’s disease. Yet robustly reproducible CSF biomarkers remain limited, raising the possibility that natural CSF protein variability itself is substantial. In this study, the authors longitudinally profiled the CSF proteome of 12 neurologically normal individuals over 4 years to quantify both intra-individual and inter-individual variation. 1) Proteomic workflow - CSF samples were analyzed by nanoLC–MS/MS after tryptic digestion, with label-free quantification using MaxLFQ - 5601 peptides and 791 protein groups were identified - Proteins detected in all 36 samples and supported by at least 2 unique peptides were retained - 216 proteins were yielded for variability analysis 2) Three types of variability: CVa, CVt, CVg - CVa = analytical coefficient of variation, estimated from repeated measurements of pooled CSF; mean 8.49% - CVt = intra-individual variability over time, calculated from repeated samples within the same subject; range 2.6%–29.3%, mean 8.2% - CVg = inter-individual variability, calculated across subjects; range 9.3%–101.5%, mean 25.7% 3) Reference change value (RCV) RCV estimates how large a change must be before it exceeds expected analytical + biological variation. The formula used was: RCV = 1.96 × √2 × √(CVa² + CVi²) where CVi represents biological variation. In this study, the authors report: - RCVt: 7.2%–81.3% - RCVg: 19.2%–281.4% This means that for some proteins, even seemingly large fold changes may still fall within the expected range of natural variation. 4) Variability of commonly proposed neurological biomarkers Several well-known candidate biomarkers showed strikingly different variability profiles: - Haptoglobin: CVg 101.5%, RCVg 281.4% - Beta-2-microglobulin: CVt 29.3%, RCVt 81.3% - APOE: CVg 31.8%, RCVg 88.2% - Clusterin: CVg 12.8%, RCVg 35.4% My scientific thought: This study shows that natural CSF protein variability is broad, both within individuals and across individuals. Inter-individual variation is especially large, reaching CVg = 101.5% in some proteins. For example, for a protein with CVg = 50%, the authors’ calculation suggests that a fold change greater than ~2.39 would be required before the difference could be considered unexpectedly large. Based on these results, the authors conclude that biomarker discovery should always evaluate protein-specific variability. This raises a further question: why are CSF proteins so variable in the first place? Approximately 80% of CSF proteins are blood-derived, and their concentrations also change along the route from the ventricles to lumbar CSF. Individual differences in permeability at the blood–brain barrier and choroid plexus may therefore create a highly selected protein population. In addition, total CSF protein concentration is only about 0.2–0.5% of that in blood, so the effects of neuronal activity, neurodegeneration, and protein release/clearance within the CNS may become relatively more visible. #Bioinformatics #Proteomics

@badlogicgames This is really smart

i don't understand half of what nico builds, but i'm always sure it is useful for many folks out there.

@mhdempsey Interested

DM me if you want to test and give feedback on a new thing i built around finding art that evokes the feelings of other art you love (think blended goodreads/letterboxd/spotify)

Notifications aren't working either :(

Love it, but it’s really slow? Some features I miss from though, like being able to self-modify. Definitely feels more secure.

@owl_posting @Blueprintbio The deployment issues with Far-UVC have been so frustrating. Incentive design e.g. tax credits seem like high ROI here.

jacob works at @blueprintbio and dramatically undersells how consequential they have been in pushing pathogen-agnostic defenses closer to the overton window. e.g.: if you're curious about far-UVC, they have published a 146 page report over it here: blueprintbiosecurity.org/u/2025/06/Blue…

Interesting enzymatic approach to clearing amyloid for AD from an Ultragenyx spinout. AAV9 delivered PPCA enzyme. Amlogenyx Announces Positive Preclinical Data on AM805, a Potent Amyloid-Degrading Protease for the Treatment of Alzheimer’s Disease amlogenyx.com/wp-content/upl…

I made a Claude Code skill that generates conference posters 🛠️ Instead of a static PDF, it outputs a single HTML file — drag to resize columns, swap sections, adjust fonts, then give your layout back to Claude. 🔁 🔗 Skill 👉 github.com/ethanweber/pos…

It was only a matter of time

one of my favorite parts of the essay. almost whimsical

@zavaindar @karpathy @ericjang11 @modal Glad I’m not the only one who played with autoresearch in pi this weekend. We’ll see if I can get this prototype working…

inspired by writings of @karpathy & @ericjang11 we built an autoresearcher via Claude & @modal within 48 hours we’d beat published baselines in protein thermostability we’re not pivoting to a neolab 🙃 but here’s what a small team w curiosity & gumption can do w todays tools

Anyone else having remote control sessions just stop connecting mid-chat on the mobile app? The sessions are still running on my laptop and get stuck "reconnecting". @bcherny