iiinova — where your digital self keeps evolving

Chris Lengerich @chrislengerich.com · 17h

The fallacy of (transitive) order: The finite action space of contrastive decision-making (a vs. b) requires an agent to impose an order on the world (1,2,3). Yet we know from watching nature that games observed honestly are often non-transitive (rock-paper-scissors).

100

Chris Lengerich @chrislengerich.com · 16/09/2026

Bumped public key fyi: www.chrislengerich.com/verified.html Part of a migration to hardware-backed E2E encryption.

chrislengerich.com

Verified Accounts

000

Chris Lengerich @chrislengerich.com · 15/09/2026

It's kinda wild that few historians have explored the different generations of workers in tech. A lot of change over the decades in the industry. (aside: it turned out "The End of History" actually meant that social media would unemploy the historians).

010

Chris Lengerich @chrislengerich.com · 15/09/2026

"How do we make scientists and engineers 100x more productive to solve problems that matter?" Wrote this about 4 years ago as a personal mission statement, I think we've more or less done it. Still, much more to come.

000

Chris Lengerich @chrislengerich.com · 23/06/2026

One of the more uncomfortable things with investing is that to make 100x, you have to be comfortable with some investments being down 50% a couple times (or worse, no updates).

100

Chris Lengerich @chrislengerich.com · 17/06/2026

5-minute meetings, and take a lot of them.

000

Chris Lengerich @chrislengerich.com · 31/05/2026

AI bubble tracker (echoes of 2022): 1. Friends are getting $100M offers to stay at bigcos. 2. Landlords raising rent 20% or more. 3. No one - investor, researcher, founder - is incentivized to short it. 4. ~9 months of runway before many startups need to raise billions, or become walking dead.

110

Chris Lengerich @chrislengerich.com · 23/05/2026

Recs for @bsky.app : 1. SEO, SEO, SEO - discoverability via search still is quite poor, even for accounts that should have domain authority for a search term. 2. Much better in-app search. 3. If using plugins for something like this, default plugins.

010

Chris Lengerich @chrislengerich.com · 23/05/2026

It's not just NLP-based coding and writing more READMEs/verification documentation for your agents, what the software goal is itself is changing based on 10x faster coding.

101

Chris Lengerich @chrislengerich.com · 21/05/2026

The abundance problem isn't really a new thing, it's actually quite old (cf. asabiyyah cycle). Just that it's happening in intelligence now vs. other products in the past. And almost as a rule, societies have not managed it well.

100

Chris Lengerich @chrislengerich.com · 14/01/2026

RIP Scott, one of the brilliant. 68 is far too young, wish you could have seen how far we'll go.

100

Chris Lengerich @chrislengerich.com · 08/01/2026

Hey @wired.com , as a umm...tech magazine, it would be wise to acknowledge your data breach ;) www.brightdefense.com/news/wired-b... Can confirm - my alias email was in the dump (notified today by a third party), and it was only used for Wired since a couple months ago.

brightdefense.com

2.3M WIRED Subscribers Exposed in Condé Nast Leak

What Happened In late December 2025, a database with 2.3+ million records tied to WIRED subscribers surfaced on underground forums. A threat actor called Lovely posted it on Breach Stars on Dec 20, cl...

000

Chris Lengerich @chrislengerich.com · 28/12/2025

Types of economic models: 1. Discriminative (distill a couple variables for a simple decision boundary, model-based) 2. Generative (distill a generative process for more complicated model, model-based)

100

Chris Lengerich @chrislengerich.com · 28/12/2025

Is something unique interesting or small? A fundamental fork that leads to a large difference in how we think about value.

000

Chris Lengerich @chrislengerich.com · 07/11/2025

Turbulence in the confused middle is now a constant feature of the new plotline (smaller variance models ftw). But you can really feel when the right plot clicks in.

100

Chris Lengerich @chrislengerich.com · 07/11/2025

Seems like it compiles and runs now. Nifty. A friend said tonight, "Thanks!" I said, "but I haven't done anything yet." "No, but you gave me hope."

000

Chris Lengerich @chrislengerich.com · 21/09/2025

Finally. While we were all looking at AI, USDC on Solana became a credible tradbank replacement. 1-2 min, $0.50 (eth) -> ~15s, <$0.01 (sol) NB: Transactions visible on chain (obv) and easy to fat-finger funds to the wrong chain (not ready for grandma yet), but still, for geeks, it just works.

100

Chris Lengerich @chrislengerich.com · 17/09/2025

Longevity metrics: Rather than talking about expected lifespan, it's probably more intuitive/useful to think about marginal cost to extend lifespan by one year. Right now, it's cheap in early age, much more expensive in old age. If we invest together, the cost will get much cheaper in the future.

000

Chris Lengerich @chrislengerich.com · 03/09/2025

Have a 10-year bet running now to never read a newsfeed in person again (using AI summarizers to stay informed and newsfeed-blocking extensions/UI hacks to enforce). Doable now, increasingly efficient over time.

100

Chris Lengerich @chrislengerich.com · 28/08/2025

Newsfeeds are probably dead (for human consumption). Replaced them with AI summaries for ~a month (scheduling utilities/async flows like Lutra/OAI helpful here). Same events, but the world feels calmer/contextualized, like reading an investor newsletter. End of an era, probably for the better.

100

Chris Lengerich @chrislengerich.com · 28/08/2025

Would be neat to see a leaderboard re: implicitly personalized AIs (those with memory-based adaptation methods) on common benchmarks? Pre-training still the biggest factor of AI intelligence by far, but how much does implicit prompting (via memory) change the intelligence between users?

000

Chris Lengerich @chrislengerich.com · 28/08/2025

Why do biotechs fail? TL;DR - scale matters, hedging matters, efficiency of a trial matters. Often the science fails first, leading to cascading management/funding failures. A lot of knobs here that can be tweaked (some obvious, others less obvious). www.librariesforthefuture.bio/p/why-do-bio...

librariesforthefuture.bio

Anatomy of a Biotech Failure

Eight Case Studies in Recent Shutdown History

010

Chris Lengerich @chrislengerich.com · 09/08/2025

Re: AI markets, would much rather see new markets for new asset classes and geos, rather than new markets for new actors (AI agents, humans w/ AI tools). Even with algos running 75% of trades in the market now, few markets have been created only *for algos.*

100

Chris Lengerich @chrislengerich.com · 09/08/2025

Typically, new ideas come from (different) fringes of the network where diverse threads start to intersect, then travel towards the center to get power (but may pick up competitors in the process).

100

Chris Lengerich @chrislengerich.com · 09/08/2025

For how similar they are in world models, VC and founders diverge quite a lot in actions. VC - horizontal, power accumulation, knows everyone. Founder - vertical, deep, iterates, knows users, team and funders. Growthco - horizontal x vertical (a la OpenAI/Google/Alibaba).

100

Chris Lengerich @chrislengerich.com · 09/08/2025

Hedging: $, portfolio, iteration, expertise, (sometimes) specialized contracts. Key insight into founders is how they react uncertainty - when you encounter uncertainty, do you raise more to hedge (portfolio hedge), or do you raise less, think more (expertise hedge)?

100

Chris Lengerich @chrislengerich.com · 08/08/2025

The Internet right now looks like a global market on the surface (anyone can read!), but it's actually a bunch of local markets disconnected from each other, unless you're already shipping at scale. What are your short growth paths to resource and user communities?

000

Chris Lengerich @chrislengerich.com · 08/08/2025

1. The economic value of prediction scales with your (real) action space. 2. What you predict about the world also depends on your (simulated) action space. Explains a lot about value systems in the modern, hyper-scale world vs. the 150-person village.

100

Chris Lengerich @chrislengerich.com · 06/08/2025

The line between investment and cognitive labor is starting to get a bit blurry. Using AI agents to do work directly, yielding $, is not that different from investing money in another human verifier to do the same thing, except ownership fraction and control (which does, of course, matter).

100

Chris Lengerich @chrislengerich.com · 28/07/2025

High-variance techs often lead to a multimodal future (not necessarily, but often).

000

Chris Lengerich @chrislengerich.com · 26/07/2025

If you can with >70% confidence, build the vertically integrated thing (definitely optimistic). Otherwise, build the personal tools that upgrade the tech era, which makes the vertically integrated thing possible, and be more indefinitely optimistic.

100

Chris Lengerich @chrislengerich.com · 17/07/2025

Arguably enough money earmarked for AI companies now that we can start talking about momentum vs. fundamental strategies in VC. True for the AI labor markets too. The music will probably stop for some big names in 1-2 years in dramatic ways, but the winners do have 10x-100x growth left.

100

Chris Lengerich @chrislengerich.com · 25/06/2025

Re-posting from Engage Bio (personal portfolio) Exciting news from Engage Bio! We’re proud to announce that the Cystic Fibrosis Foundation has made an equity investment in Engage to support development of durable, non-viral genetic medicines for cystic fibrosis.

linkedin.com

Exciting news from Engage Bio! | Will Olsen

Exciting news from Engage Bio! We’re proud to announce that the Cystic Fibrosis Foundation has made an equity investment in Engage to support development of durable, non-viral genetic medicines for c...

110

Chris Lengerich @chrislengerich.com · 23/06/2025

Objectively valuable milestones - a conversation that comes up again and again in hardtech. Hardtech cos/intrinsically motivated scientists often have complex intermediate goals and unique theory. The market, though, is (on average), not enlightened and not patient. youtu.be/aG8_AeiwDPA?...

youtu.be

Celine Halioua: The Key to Longevity

YouTube video by S3

100

Chris Lengerich @chrislengerich.com · 04/06/2025

Investment learnings from the last few years: 1. Contrastive distillation works better than most theories as a model of reasoning and writing. It's simple, it scales. 2. But data/experience increases precision. Hang out near the reefs. PhD-led startups with good coaches do unreasonably well.

100

Chris Lengerich @chrislengerich.com · 04/06/2025

Only 24 hours in a day - focusing on flow with portcos and family matters more than one might think.

000

Reposted by Chris Lengerich

Cancer Research UK @cancerresearchuk.org · 10/04/2025

Scientists we co-fund at @CambridgeUni.Bsky.Social have developed an algorithm that could pave the way for more personalised cancer treatments! It can pinpoint patterns of mutation in cancer cells and could help identify the patients who could benefit most from immunotherapy. Read more 👇

cruk.ink

Scientists develop a 'metal detector' to find cancer's weak points - Cancer Research UK - Cancer News

PRRDetect, which was developed with our funding, picks out cancers that are vulnerable to immunotherapy by scanning for mutation patterns.

2177

Chris Lengerich @chrislengerich.com · 29/03/2025

Naive, high-level take on aging - we understand some of the key variables, but not yet the regulatory dynamics. You need these for precise control (and therapeutics that pass clinical trials).

100

Reposted by Chris Lengerich

NYAA Starfest @nyaastarfest.bsky.social · 05/02/2025

IC 410 is located 12,000 lightyears from Earth in the Auriga constellation, and is nicknamed the Tadpole Nebula because of the tadpole-shaped clouds of dark dust that appear to be swimming towards the centre. #astrophotography . I took this last night with HA,S2 and O3 filters.

0323

Reposted by Chris Lengerich

Randall Munroe @xkcd.com · 26/02/2025

RNA xkcd.com/3056

154177422542

Chris Lengerich @chrislengerich.com · 21/12/2024

Regulatory genomics is highly underinvested. May unlock cures to aging within reasonable timespans, starting to get dense predictive data (w/ single-cell datasets), contrastive distillation can help in sorting the information dynamics out. All quite early, but promising.

120