Paul Sample @paulsample.bsky.social · 12/02/2026At what point do you decide to just not publish a paper because the models are so out of date? This paper was published yesterday but uses models that are ancient by today's standards (GPT 4, Gemini 2, Llama 2 & 3). 100
Paul Sample @paulsample.bsky.social · 01/09/2025I think these guys are based in the US and offer GLM-4.5 synthetic.new/landing/homesynthetic.newSynthetic | Run LLMs, privatelyChat with open-source models privately 032
Paul Sample @paulsample.bsky.social · 22/08/2025openai.com/index/accele...openai.comAccelerating life sciences researchOpenAI and Retro Biosciences achieve 50x increase in expressing stem cell reprogramming markers. 050
Paul Sample @paulsample.bsky.social · 21/08/2025cloud.google.com/blog/product...cloud.google.comMeasuring the environmental impact of AI inference | Google Cloud BlogA methodology for measuring the energy, emissions, and water impact of Gemini prompts shines a light on the environmental impact of AI inference. 010
Paul Sample @paulsample.bsky.social · 08/07/2025Today would be a great day for the AI research community to leave X for good. In your own self interest and for the public perception of AI, take a stand. Unless you agree with using it as a misinformation and propaganda machine, I guess. 020
Reposted by Paul SampleAi2 @ai2.bsky.social · 01/07/2025Introducing SciArena, a platform for benchmarking models across scientific literature tasks. Inspired by Chatbot Arena, SciArena applies a crowdsourced LLM evaluation approach to the scientific domain. 🧵 1155
Paul Sample @paulsample.bsky.social · 12/06/2025Anthropic's Claude Code Python SDK just released? Currently sitting at 1 star! github.com/anthropics/c...github.comGitHub - anthropics/claude-code-sdk-pythonContribute to anthropics/claude-code-sdk-python development by creating an account on GitHub. 011
Reposted by Paul SampleAI x Bio Discovery @aixbiobot.bsky.social · 20/05/2025Robin: A multi-agent system for automating scientific discovery [new] Automating scientific discovery via a multi-agent system that integrates literature search and data analysis to propose/validate hypotheses. 001
Paul Sample @paulsample.bsky.social · 16/05/2025The learning objective is next token prediction, but to do that well the models have to learn a deeper representation of “the world”. Compare the latest LLMs to a simple N gram model and tell me that they’re just “next word” predictors. 140
Reposted by Paul SampleNicole Hennig @nic221.bsky.social · 14/05/2025Power-Hungry Data Centers Are Warming Homes in the Nordics bloom.bg/4jYABDI #AI #DataCenters #climate 043
Paul Sample @paulsample.bsky.social · 14/05/2025WhyTF would people use Grok when they do things like this? Not to mention the pollution they’re producing in Memphis. 010
Paul Sample @paulsample.bsky.social · 13/05/2025There is no bigger turn off to reading a paper than seeing the words “is all you need” in the title. 110
Paul Sample @paulsample.bsky.social · 10/05/2025Finally, the Twitter/X killer. Nothing quite as satisfying as operating in purely in a terminal. www.itter.shitter.shitter.sh - Social Media via SSHitter.sh: Ironic, text-only, SSH-based social networking for terminal lovers. No browser, no js, just eets. 000
Paul Sample @paulsample.bsky.social · 02/05/2025I always look forward to FutureHouse releases. I had to do a little digging for API information so here it is for those who are interested. futurehouse.gitbook.io/futurehouse-...futurehouse.gitbook.ioFutureHouse Platform API Documentation | FutureHouse Cookbook 041
Reposted by Paul SampleAngie Rasmussen @angierasmussen.bsky.social · 30/04/2025I knew he would do this. I wrote a whole article about how he would do this. I am not surprised he’s undermining vaccination. I know he’s an anti-vaxxer & a predatory crank. And yet…you have got to be fucking kidding me. He’s promoting infant mortality. 17644193
Paul Sample @paulsample.bsky.social · 30/04/2025We have to be reaching the sycophancy floor soon.. right? 020
Paul Sample @paulsample.bsky.social · 19/04/2025I’m curious about what your thoughts are on the odds of Trump et al putting a finger on the scale if Tesla shares tank. And ultimately, what could they reasonably do to give it some juice. 110
Reposted by Paul SampleAI x Bio Discovery @aixbiobot.bsky.social · 24/02/2025Tahoe-100M: A Giga-Scale Single-Cell Perturbation Atlas for Context-Dependent Gene Function and Cellular Modeling [new] Atlas maps 50 cancer lines' trancriptomic response to 1.1k mols, aiding AI modeling. 111
Reposted by Paul SampleAI x Bio Discovery @aixbiobot.bsky.social · 22/02/2025Genome modeling and design across all domains of life with Evo 2 [new] Evo 2: Learns genomic features across life's domains from DNA alone and generates coherent, controllable genome sequences at scale. 021
Paul Sample @paulsample.bsky.social · 21/02/2025"Most baffling to many were the firings of hundreds whose jobs were not funded by taxpayers. Their positions were financed through congressionally approved agreements that routed fees from the drug, medical device and tobacco industries to the agency." 🤡 www.nytimes.com/live/2025/02...nytimes.comTrump Live Updates: Layoffs, Ukraine War and Deportation News 020
Reposted by Paul SampleKaren Hao @karenhao.bsky.social · 21/02/2025For decades, the US government has painstakingly kept American science #1 globally—and every facet of American life has improved because of it. The internet? Flu shot? Ozempic? All grew out of federally-funded research. Now all that's being dismantled. 1/ www.technologyreview.com/2025/02/21/1...technologyreview.comThe foundations of America’s prosperity are being dismantledFederal scientists warn that Americans could feel the effects of the new administration's devastating cuts for decades to come 7829991497
Paul Sample @paulsample.bsky.social · 04/02/2025Has anyone tried this? Struggling to find an information retrieval system I like. www.perplexity.ai/help-center/...perplexity.aiWhat is Internal Knowledge Search | Perplexity Help CenterInternal Knowledge Search allows both Perplexity Pro and Enterprise Pro users to search across internal files alongside the web. 050
Reposted by Paul SampleSimon Willison @simonwillison.net · 31/01/2025Severance obsessives like myself may be interested to hear that the first 8 chapters of The You You Are by Dr. Ricken Lazlo Hale, PhD are now available for free as both text and an audiobook read by Ricken himself: books.apple.com/us/book/the-...books.apple.comThe You You AreHealth, Mind & Body · 2025 2131
Paul Sample @paulsample.bsky.social · 27/01/2025Avoidance is better for mental health but not for decision making in an increasingly complex and unpredictable world. 050
Reposted by Paul SampleSung Kim @sungkim.bsky.social · 24/01/2025Aider reports that R1+Sonnet (R1 Thinking + Sonnet) set a new SOTA on the aider polyglot benchmark at 14X less cost compared to o1. 64% R1+Sonnet 62% o1 57% R1 52% Sonnet 48% DeepSeek V3 aider.chat/2025/01/24/r...aider.chatR1+Sonnet set SOTA on aider’s polyglot benchmarkR1+Sonnet has set a new SOTA on the aider polyglot benchmark. At 14X less cost compared to o1. 1419
Paul Sample @paulsample.bsky.social · 21/01/2025I tried a few biology-related queries and was impressed with the results. Researchers should check this out. 0110
Paul Sample @paulsample.bsky.social · 18/01/2025How have I been coding with Python for 10 years and just learned today that you can update the same line in a terminal using '\r' and end=''? print(f"\rDownload progress: {percent}%", end='') Thanks Claude. 0122
Reposted by Paul SampleEric Topol @erictopol.bsky.social · 16/01/202598% of microplastic particles (MP) in bottled water are smaller than 20 μm, which are much more likely to get thru the GI tract to blood and organ tissue. Municipal tap water was higher in MP than 8 of 10 bottled water brands journals.plos.org/water/articl... 1223474
Reposted by Paul SampleVikram Agarwal @vagar.bsky.social · 15/01/2025Super excited to announce our latest work. On a personal note, it's not an exaggeration to say that blood, sweat, and tears got us to the finish line on this: working w/ an outstanding global team of scientists in Germany, Japan, Russia, and USA responding in >100 pages of complex reviewer comments.nature.comMassively parallel characterization of transcriptional regulatory elements - NatureLentivirus-based reporter assays for 680,000 regulatory sequences from three cell lines coupled to machine-learning models lead to insights into the grammar of cis-regulatory elements. 23610
Paul Sample @paulsample.bsky.social · 09/01/2025Gemini 2.0 Flash gets straight to the point with "peanut". 110
Paul Sample @paulsample.bsky.social · 08/01/2025Not to brag but apparently I'm "on a roll!" over at LinkedIn. Blessed 🙏. 050
Reposted by Paul SampleJeremy Howard @howard.fm · 19/12/2024I'll get straight to the point. We trained 2 new models. Like BERT, but modern. ModernBERT. Not some hypey GenAI thing, but a proper workhorse model, for retrieval, classification, etc. Real practical stuff. It's much faster, more accurate, longer context, and more useful. 🧵 19620147
Reposted by Paul SampleDavid Bikard @dbikard.bsky.social · 13/12/2024Excited to share that we have just released genomenotebook 1.0 together with @trichomedoctor.bsky.social ! If you like to analyze genomics data in Jupyter notebooks this is for you. You can get beautiful genome browsers directly in your notebooks: dbikard.github.io/genomenotebo...dbikard.github.iogenomenotebookA genome browser in your Jupyter notebook 37722
Paul Sample @paulsample.bsky.social · 11/12/2024I know this store. It's in the well-to-do Magnolia neighborhood of Seattle. It's a little rundown but otherwise fine.The sign describes a plan to demolish it in order to build apartments (or something) but there has been no movement on it in years. 000
Paul Sample @paulsample.bsky.social · 08/12/2024Nice tool and the UI is great. It would be helpful if you could filter for specific file types. For example, ingesting github.com/google-gemin... pulls out tons of stuff you wouldn't actually need for an LM. 110
Reposted by Paul SampleRom2 @yasbaltrine.bsky.social · 07/12/2024I made this free and open source dev tool: gitingest.com Replace "hub" with "ingest" in any github URL to get a text digest that you can feed into any LLM 167615
Paul Sample @paulsample.bsky.social · 08/12/2024Great, and for the pay-as-you-go tier the experimental 1.5 models are charged at the rates of 1.5 pro? 000
Paul Sample @paulsample.bsky.social · 08/12/2024Very cool. I know that the experimental models are free of charge in AI Studio, but do you know if that’s true when using the API? I can’t seem to find a clear answer. 110
Paul Sample @paulsample.bsky.social · 05/12/2024Once again reminded of the amazing storytelling through data visualization that W.E.B Du Bois was doing in 1899! 040
Reposted by Paul SampleDiego del Alamo @delalamo.xyz · 04/12/2024New ESM models! Unlike ESM3 these seem to be classic sequence-only PLMs. The 300M and 600M param models are open source and available for commercial use, while the 6B param model is academic use only www.evolutionaryscale.ai/blog/esm-cam... 1204
Paul Sample @paulsample.bsky.social · 01/12/2024However it is not deterministic! I've rerun the same abstracts and noticed some borderline papers get classified differently with each run. 040
Paul Sample @paulsample.bsky.social · 01/12/2024I'm really enjoying the structured output ability of Gemini for this zero shot classification problem. I'm trying to categorize bioRxiv papers to focus on the ones most relevant to my interests and it's working with very little fiddling. 1110