iiinova — where your digital self keeps evolving

Abhishek Jha @ajelucidata.bsky.social · 06/10/2026

The original Turing test: can a machine convince you it's human? CAPTCHAs: can a human convince a machine it's not a bot? We prove we're real by being slow and bad at identifying crosswalks. Somewhere Alan Turing is amused.

000

Abhishek Jha @ajelucidata.bsky.social · 05/10/2026

Biological evidence can mislead systems in a quieter way than typos or acronyms. In Open Targets’ AML evidence, ~31% of rows come from paper introductions and discussions - not the sections reporting new findings. A co-occurrence score can’t tell background from evidence.

100

Abhishek Jha @ajelucidata.bsky.social · 28/09/2026

A small data issue can quietly corrupt biological AI. We looked at literature evidence for AML in Open Targets. ~209K of 884K rows matched “MS” and were classified as myeloid sarcoma. But after checking the actual papers, only 430 were confirmed as myeloid sarcoma. That’s 0.2%.

110

Abhishek Jha @ajelucidata.bsky.social · 23/09/2026

Every few weeks now, someone sends me a new #AI announcement and asks a version of the same question: are you about to get eaten? It’s a fair question, and I sat down to answer it honestly in my latest newsletter. You can read it here: substack.com/home/post/p-...

substack.com

When Reasoning Becomes Abundant, Evidence Becomes Scarce

The durable layer may sit below the model.

000

Abhishek Jha @ajelucidata.bsky.social · 21/09/2026

Last week we hosted the 9th edition of #DataFAIR in Cambridge, and I was quite happy with how the day turned out :) Thank you to everyone who brought their time and perspective. 9 editions in, and what started as an event feels much more like a community now.

000

Abhishek Jha @ajelucidata.bsky.social · 17/09/2026

#DataFAIR, our flagship event, returns today for its 9th edition! An afternoon with leaders from Takeda, Pfizer, Compass Therapeutics, Paratus Sciences and others, on reshaping early discovery and moving computational biology into real R&D decisions. See you in Cambridge :)

000

Abhishek Jha @ajelucidata.bsky.social · 16/09/2026

I wrote a new piece for #DrugDiscoveryNews on why “more data” isn’t necessarily the answer to better drug discovery. We have more biological associations and connections than ever, yet R&D productivity continues to decline. You can read the article here: www.drugdiscoverynews.com/we-have-more...

drugdiscoverynews.com

We have more biological data than ever. Are we asking the right questions? | Drug Discovery News

We have never had more biological data, or more ways to connect it. The programs that win will be the ones that ask better questions of it, not bigger ones.

010

Abhishek Jha @ajelucidata.bsky.social · 15/09/2026

I am attending #ddpwest2026 on October 15. Excited for discussion around the #AI era, and especially interested in talks, panels, and roundtable around the Inside The Business track topics. Looking forward to joining Jin Hyun J.'s session! Thanks #DataDrivenPharmaWest for hosting!

000

Abhishek Jha @ajelucidata.bsky.social · 10/09/2026

AI that reads only abstracts misses more than you’d think. We built knowledge graphs from 21,926 #Alzheimers papers using Wiley’s corpus. Full text surfaced 11.4% more connections and 7.2% more concepts. Better inputs mean better AI. Read the full study here: content.wiley.com/showyourwork...

000

Abhishek Jha @ajelucidata.bsky.social · 09/09/2026

How many tokens I can spend dictate my #work-life balance. #2026

000

Abhishek Jha @ajelucidata.bsky.social · 08/09/2026

A quick trip to San Diego last week! The picture is from one of our small group dinners, something we host often. Met old friends and new faces. What I love about these dinners is the different perspectives I get to hear. If you’d like to join one sometime, DM me. I’d be glad to have you :)

010

Abhishek Jha @ajelucidata.bsky.social · 26/08/2026

Upcoming panel discussion! Why 2 patients with same diagnosis respond so differently to the same therapy? On Sep 1, Fred Baribaud, Hozefa Bandukwala & Sreeram Balasubramanian will discuss where the field is making progress,where it still falls short & what could shape next gen of I&I therapies.

100

Abhishek Jha @ajelucidata.bsky.social · 25/08/2026

I wrote a piece for Drug Target Review questioning whether bigger foundation models are enough for single-cell biology. They’ve transformed protein structure & weather prediction. But in single-cell, scale has failed to deliver the results many expected. So what’s actually holding us back? Link 👇

100

Abhishek Jha @ajelucidata.bsky.social · 13/08/2026

11 years to team EL. We started by helping scientists spend less time on messy data & more on science, I hold that idea a little more humbly than I used to. Today, 90+ biopharma R&D teams trust us, & 40+ prog we’ve supported have moved beyond IND. Grateful to everyone ❤️

000

Abhishek Jha @ajelucidata.bsky.social · 09/07/2026

For most of my career, I leaned heavily on public data for many scientific problems. RCSB.org, GEO, NCBI, and others. You search everything you can think of, you convince yourself you have been thorough, and you are still left with the quiet certainty that something is out there you did not find.

210

Abhishek Jha @ajelucidata.bsky.social · 22/06/2026

I woke up this morning to find that our AI Lab team had published Elucidata’s PubMed MCP. As a recovering scientist, I did the most predictable thing. A cup of coffee & a few 1k tokens later, I had a quick benchmark of our MCP server against the Claude marketplace PubMed MCP

230

Abhishek Jha @ajelucidata.bsky.social · 19/06/2026

Most target discovery teams are asking the wrong question. “What is associated with this disease?” is now table stakes. A good knowledge graph can surface it. So can a well-scaffolded LLM workflow.

210

Abhishek Jha @ajelucidata.bsky.social · 18/06/2026

Heading to BIO International Convention 2026 in San Diego next week. Always one of the better weeks of the year for sitting down in person with leaders across #biotech & #pharma.

100

Abhishek Jha @ajelucidata.bsky.social · 13/06/2026

Elucidata will turn 11 this year. I have been thinking about the lessons that took the longest to learn. One of those is that the hard problem is almost never the technology. As scientists, we are wired to believe it is. #podcast

210

Abhishek Jha @ajelucidata.bsky.social · 08/05/2026

How many tokens I can spend dictate my work-life balance. #AI #AI2026

100

Abhishek Jha @ajelucidata.bsky.social · 23/04/2026

110

Abhishek Jha @ajelucidata.bsky.social · 20/04/2026

We're at AACR in San Diego this week, presenting our work on neuroendocrine prostate cancer. NEPC shows up in up to 30% of CRPC patients. Median survival is about 7 months.

100

Abhishek Jha @ajelucidata.bsky.social · 02/03/2026

We benchmarked Biomni, SpatialAgent, and Polly on spatial transcriptomics. Results: "Precision" math is useless if the citations are fabricated. We favor traceability over flashy, unverified outputs. Reliability > Hype. Full deep dive: www.elucidata.io/blog/compara... #HealthcareAI #Drugdiscovery

elucidata.io

Comaparative analysis of Stanford’s Biomni, Genentech’s SpatialAgent, and Polly Bioagent

This blog evaluates how autonomous AI handles spatial transcriptomics, cell-cell communication, and literature-based drug repurposing, highlighting critical differences in algorithmic rigor and factua...

110

Abhishek Jha @ajelucidata.bsky.social · 28/01/2026

A three‑hour “boring” deep dive on a hospital software company shouldn’t be the thing that shakes your worldview, but this one did for me. Epic went from “annoying EMR vendor” in my head to a quiet masterclass in vertical software, data moats, and playing the long game in a life‑or‑death domain.

110

Abhishek Jha @ajelucidata.bsky.social · 14/01/2026

AI is great at pattern matching. In drug discovery, that’s not enough. Joined the @DataStorage team to talk about: Why pattern‑matching AI breaks in messy biology OOD problems as the real opportunity Data‑centric vs model‑centric AI Infra, costs, and deploying safely in regulated environments

210

Abhishek Jha @ajelucidata.bsky.social · 30/12/2025

Not a victory lap. Just an honest check-in on what moved, what hurt, and what’s worth doubling down on next year. This newsletter is a simple pass through my own highlights, lowlights, and funlights from 2025, and a small prompt to choose better inputs, not just chase better outcomes, in 2026.

100

Abhishek Jha @ajelucidata.bsky.social · 08/11/2025

Every biopharma leader agrees: “Pre-competitive data sharing is essential.” Yet when it’s time to act, no one moves first. Here's what I think about it- 👇

100

Abhishek Jha @ajelucidata.bsky.social · 07/11/2025

Some people think “data-centric AI” means having more data. It doesn’t. It means..... ..Having fewer regrets! #CleanData #SmarterAI #DataCentricAI

000

Abhishek Jha @ajelucidata.bsky.social · 04/11/2025

Most founders over-index on product. GTM success = knowing your ACV, product complexity, and who decides (champion or committee). “AI” isn’t a GTM differentiator, clear value is. What GTM mistake slowed you down the most? #startups #AI #healthtech

011

Abhishek Jha @ajelucidata.bsky.social · 16/10/2025

Gearing up for 𝗗𝗮𝘁𝗮𝗙𝗔𝗜𝗥 2025 in Boston. 7 years in, we ask the hard questions: • Can AI & LLMs actually speed drug discovery? • What bold bets will reset pharma? • How do we move from pilots → production? Oct 23, let’s dig in. #LifeSciences #AIinBiopharma

110

Abhishek Jha @ajelucidata.bsky.social · 15/10/2025

The "we'll build our own AI" era just ended. Takeda ↔ Nabla Bio: Renewed Sanofi: New AI deal Salt AI: $10M raised Shuttle acquires Molecule. ai The smart money figured out: AI isn't a feature you build—it's an ecosystem you join.

110

Abhishek Jha @ajelucidata.bsky.social · 23/09/2025

Earlier last week in #Philadelphia, we hosted a small group dinner during #BioTechX USA 2025. It felt great to sit down with heads of R&D, data, and operations from #pharma, #biotech, and #diagnostics hearing directly what they care about and what still trips teams up.

110

Abhishek Jha @ajelucidata.bsky.social · 12/09/2025

Clinical & real-world datasets can unlock breakthroughs in #drugdevelopment & #precisionmedicine - but patient data is fragmented across labs, imaging, omics & outcomes. Join us to see how we unify it into a #KnowledgeGraph! 👉 bit.ly/kg-webinar_p...

010

Abhishek Jha @ajelucidata.bsky.social · 04/09/2025

Data scientist: “Do we need provenance tracking?” "Yes. Because blind trust is not FDA-compliant."

000

Abhishek Jha @ajelucidata.bsky.social · 01/09/2025

One of my personal goals has always been to give back time to scientists - so they can focus on what they do best: bringing life-changing therapies to patients. This goal didn’t come from a business plan. It came from a conversation that’s stayed with me for years. 1/4

100

Abhishek Jha @ajelucidata.bsky.social · 31/08/2025

One of my biggest goals has always been simple: give scientists back time. Because behind every therapy, every discovery, is someone fighting not just biology, but time. And sometimes an extra hour can change everything. #personalstory #CEO

000

Abhishek Jha @ajelucidata.bsky.social · 29/08/2025

“A team of PhD-level experts in your pocket.” That’s how Sam Altman pitched #GPT5 We stress-tested that claim on a real #genomics problem - the kind scientists wrestle with daily in code and data. Did it live up to the hype? 👉 elucidata.io/blog/gpt5-ph... #PhD #GPT5

000

Abhishek Jha @ajelucidata.bsky.social · 25/08/2025

Delays in PK/PD data retrieval from trial PDFs can stall decisions by up to 2 weeks, push trial start dates back by months & even give competitors early access to trial sites. Time to rethink data accessibility in clinical research. 👉 bit.ly/webinar-xtract #ClinicalTrials #Discovery #AI

010

Abhishek Jha @ajelucidata.bsky.social · 07/08/2025

No one announces a “data quality failure.” It shows up as delays, poor model performance, regulatory headaches, and slipping timelines. We’re great at collecting data, less so at making it usable. In science, shaky data is a risk, not just tech debt. #Biomedicaldata #AI

001

Abhishek Jha @ajelucidata.bsky.social · 06/08/2025

I recently spoke to a biotech team scaling translational research. They have it all - EHRs, omics, imaging, trial data. On paper: perfect. In reality: chaos. Mismatched timestamps, missing metadata, systems that don’t talk. The data exists. But it’s not usable. #dataquality #AI

001

Abhishek Jha @ajelucidata.bsky.social · 04/08/2025

1 day left. Have you registered yet? 🔗 Register: bit.ly/kg_webinar #webinar #AI #knowledgegraph #research

011

Abhishek Jha @ajelucidata.bsky.social · 31/07/2025

🔗 Register: bit.ly/kg_webinar Polly KG introduces a programmable knowledge layer purpose-built for life sciences R&D. It supports natural language queries, cross-species reasoning, and domain-specific scoring within secure, enterprise environments.

100

Abhishek Jha @ajelucidata.bsky.social · 28/07/2025

After my recent post, it’s clear: the pain around knowledge graphs in biopharma is real. Disconnected schemas. No context. Engineers overloaded for basic queries. We’ve been there. We’ve felt that too. And it’s exactly why we built Polly KG.

100

Abhishek Jha @ajelucidata.bsky.social · 25/07/2025

When we built Polly KG, we weren’t trying to make just another #knowledgegraph. We built it to move past rigid schemas, surface NLP, and endless engineering loops - so scientists could just ask and answer questions that matter. Join us on Aug 5 - bit.ly/kg_webinar #webinar #AI #bioinformatics

100

Reposted by Abhishek Jha

Abhishek Jha @ajelucidata.bsky.social · 17/07/2025

We recently ran a closed Agentic AI workshop with biopharma R&D leaders - every seat filled. Now we’re opening it up virtually. Join us to build AI agents that actually work with biomedical data. Request your invite: bit.ly/AgenticAIwor... #AI #biopharma #clinical #workshop

013

Abhishek Jha @ajelucidata.bsky.social · 20/07/2025

Have you registered yourself yet? If not, you should do it soon. 👉 bit.ly/AgenticAIwor... #AI #data #workshops #AgenticAI

001

Abhishek Jha @ajelucidata.bsky.social · 18/07/2025

Still thinking about the energy at the Festival of Genomics & Biodata in Boston! Every R&D conversation circled back to AI in cohorting, multi-omics, and the need for better, context-rich data. Thanks to all who joined our session or stopped by. Snapshots from the talk👇

000

Abhishek Jha @ajelucidata.bsky.social · 17/07/2025

013

Abhishek Jha @ajelucidata.bsky.social · 16/07/2025

Join us tomorrow at 10:30 AM PT to learn what AlphaGenome delivers and how to apply its predictions effectively in your workflows. #AI #Genomics #Medicine Register: bit.ly/alphagenomew... Last day to join us!

001

Abhishek Jha @ajelucidata.bsky.social · 16/07/2025

Everyone built data lakes. Now we’re drowning in unusable data. AI doesn’t need lakes. It needs pipelines -focused, governed, built for decisions. The real bottleneck? Not compute. Not models. It’s data usability. Don’t pour more. Build better. #Dataproducts #AI #dataquality

000