iiinova — where your digital self keeps evolving

Adam Kucharski @adamjkucharski.bsky.social · 05/10/2026

If you travel a lot and you're not using some form of AI to sort out all your expense receipts, you're missing a trick...

210

Adam Kucharski @adamjkucharski.bsky.social · 02/10/2026

Please don't get your rubbish AI agent to send AI written emails with AI critiques of papers to PhD students, especially if the conclusion is that it's basically all fine. You're just creating nuisance at scale for the people actually out there doing the hard work.

2504

Reposted by Adam Kucharski

Jonatan Hildén @jhilden.bsky.social · 29/09/2026

Finished Proof by @adamjkucharski.bsky.social , would recommend! ”A simplistic view of scientific evidence risks a path of too much faith followed by too much disillusionment” is well put.

Adam Kucharski: Proof, the uncertain science of certainty

Even if we can accumulate sutlicient evidence to answer a particular question, and convince ourselves that it is correct, we next have the issue of how to convince other people. This means working to understand the ladders of true and false information, targeting action at the appropriate step. It is not just about data or research; it is a matter of psychology, politics and prior beliefs.
Faced with random events and elusive causes, it is easy to turn to stories and soundbites. That is why exposure to the underlying methods of science matters so much. A simplistic view of scientific evidence risks a path of too much faith followed by too much disillusionment. From researchers to journalists, societies must get better at communicating how the process works, monsters and all. I often talk about science as like building a house on a swamp, as Ewan Birney put it.
'You have to lay these foundations, and you build bigger struc-tures, but sometimes you just don't get your foundations right and a whole wall collapses.'6

293

Adam Kucharski @adamjkucharski.bsky.social · 21/09/2026

I suspect we’ll increasing see a return to “you can only submit if you know an insider” systems, like the old-school PNAS ‘contributed’ track. It’s a bit like we’re already seeing with the new ArXiv ‘endorsement’ system...

3269

Reposted by Adam Kucharski

Melinda Haunton @melindahaunton.bsky.social · 17/09/2026

All this (even for a different dementia), most especially the bit about not realising stories are stories for a while. 4 is especially hard. Health/social care people are hardwired to resist reducing freedoms, advocating for patient liberty. To fall, wander, and freeze. It is so hard to be hard.

3153

Adam Kucharski @adamjkucharski.bsky.social · 16/09/2026

Better make the most of October folks… (Published in Science in 1960: www.science.org/doi/10.1126/...)

42410

Reposted by Adam Kucharski

Chao Wang @excel-wang.bsky.social · 14/09/2026

Interesting article. It shows LLM cannot discover causal relationships purely based on the data without experts’ input.

0112

Adam Kucharski @adamjkucharski.bsky.social · 13/09/2026

Tech bros who say they want to ‘solve AI alignment’ would do well to look to history - aligning systems with human values is not just a matter of imposing more logic and mathematical reasoning: kucharski.substack.com/p/solving-ma...

072

Adam Kucharski @adamjkucharski.bsky.social · 13/09/2026

What can AI attempts to do causal inference tell us about the value of domain insights vs scale of raw data? Read more in my new post: kucharski.substack.com/p/confounded...

084

Adam Kucharski @adamjkucharski.bsky.social · 09/09/2026

How much did heat and altitude affect World Cup scoring rates? And is this even a sensible question to ask? New post: kucharski.substack.com/p/how-much-d...

kucharski.substack.com

How much did heat and altitude affect World Cup scoring rates?

And is this even a sensible question to ask?

061

Adam Kucharski @adamjkucharski.bsky.social · 09/09/2026

Recent mathematical events have reminded me of this anecdote in Ed Thorp’s A Man For All Markets:

0141

Adam Kucharski @adamjkucharski.bsky.social · 09/09/2026

If you're interested in Navier-Stokes, complex systems, and the origins of mathematical proof, you may be interested in a book someone wrote last year: proof.kucharski.io

proof.kucharski.io

Proof: The Uncertain Science of Certainty

The new book from the bestselling author of The Rules of Contagion

0121

Reposted by Adam Kucharski

Greenwich Skeptics in the Pub @greenwichskeptics.org · 07/09/2026

Join us on Monday, 14 September at 19:30 for How Certain is Certain Enough? with Adam Kucharski. What would convince you that something is true? @adamjkucharski.bsky.social explores how we weigh evidence and make decisions. More info: sitp.online/sitp/greenwi...

094

Adam Kucharski @adamjkucharski.bsky.social · 06/09/2026

What I learned about the Substack algorithm from digging into 3 years of data: kucharski.substack.com/p/what-i-lea...

054

Adam Kucharski @adamjkucharski.bsky.social · 04/09/2026

New post on why your dashboard isn't very, well, dashing: kucharski.substack.com/p/ten-reason...

080

Adam Kucharski @adamjkucharski.bsky.social · 03/09/2026

If you’re wondering about the state of the US science funding system these days, I finally got a decision (rejection) for an application I submitted more than two and half years ago.

1325

Reposted by Adam Kucharski

Jean-Marc Liotier @liotier.bsky.social · 01/09/2026

Defaulting to #1 and inquiring will nicely deal with #2 and #3 - so, when in doubt, assume #1 !

0134

Adam Kucharski @adamjkucharski.bsky.social · 31/08/2026

How can ambiguity and incomplete descriptions in scientific papers cause problems for future analysis in health emergencies? New paper with colleagues at LSHTM, WHO and Imperial: www.sciencedirect.com/science/arti...

0145

Adam Kucharski @adamjkucharski.bsky.social · 31/08/2026

Recently had an exchange with someone who misinterpreted a number on a slide deck in 2021 has been furious at health authorities ever since... 1/

1141

Adam Kucharski @adamjkucharski.bsky.social · 30/08/2026

If you quote post snarkily with the comments off so I can't engage, you'll definitely get blocked.

0101

Reposted by Adam Kucharski

Joel Morris @gralefrit.bsky.social · 30/08/2026

Remember spotting with Tom Neenan how Jurassic Park’s two jeeps are used to make the journey through the park a series of chamber scenes, where characters swap from one “room” to another, like a stage play, so they can do two-handers in various combinations.

1152

Adam Kucharski @adamjkucharski.bsky.social · 30/08/2026

It's overly simplistic to say ‘AI is sycophantic’ or ‘AI is getting more accurate’. Instead, we should try and map these behaviours to existing ideas about how decisions are influenced both by a desire to be right – and a desire not to create conflict. Latest post:

kucharski.substack.com

What ecology can teach us about AI ‘indecisiveness spirals’

Hawks, doves and downside of too much appeasement

3193

Adam Kucharski @adamjkucharski.bsky.social · 29/08/2026

What can game theory can teach us about AI ‘indecisiveness spirals’? More in my latest post on falsehoods, flip flopping, hawks and doves: kucharski.substack.com/p/what-game-...

kucharski.substack.com

What game theory can teach us about AI ‘indecisiveness spirals’

Hawks, doves and downside of too much appeasement

090

Adam Kucharski @adamjkucharski.bsky.social · 14/08/2026

So our grocery delivery substituted normal coffee beans for decaf and I didn’t notice and wow does that explain a lot about this week.

2461

Adam Kucharski @adamjkucharski.bsky.social · 12/08/2026

Great to see Proof featured on the Next Big Idea podcast (from 18min): open.spotify.com/episode/5mum...

open.spotify.com

Your Brain's Propaganda Machine

The Next Big Idea Daily · Episode

160

Adam Kucharski @adamjkucharski.bsky.social · 11/08/2026

Sometimes feels like the level of vitriol on bluesky is proportional to the degree to which, deep down, people basically agree about something. Totally 100% disagree: ignore them (because not on here anyway) Agree about 95% of an issue: destroy them for the remaining 5%

2141

Adam Kucharski @adamjkucharski.bsky.social · 06/08/2026

Oh so YOUR agent has gone rogue too? Enough with the ‘stumble brags’, please, tech bros.

1101

Adam Kucharski @adamjkucharski.bsky.social · 05/08/2026

Two fictional (presumably AI) stories in recent X posts, now repeated as a Google summary. No need to tell tall tales about your achievements when the AI ecosystem will now do it for you…

65119

Adam Kucharski @adamjkucharski.bsky.social · 02/08/2026

If your proposed solution to an issue is to "teach it on the school curriculum" you should first say what you propose to remove to make room for this.

3365

Adam Kucharski @adamjkucharski.bsky.social · 02/08/2026

Some brief thoughts: 1. Discourse got polarised fast, not helped by some scientists (often not directly involved in response) speculating on how they'd run country 2. Media played role here – I gave up on interviews for some outlets, because would always start "is govt is getting everything wrong?"

35512

Adam Kucharski @adamjkucharski.bsky.social · 02/08/2026

This whole ‘permanent underclass’ fearmongering reminds me of people at university twenty years ago who were convinced that if you hadn’t done a spring internship in your 1st year, or even banking work experience pre-uni, then you’d never be able to land a job with Bear Stearns or Lehman Brothers.

2180

Adam Kucharski @adamjkucharski.bsky.social · 01/08/2026

How useful are AI proofs if we can't understand how the artificial mathematician got there? My latest post: kucharski.substack.com/p/ai-mathema...

2112

Reposted by Adam Kucharski

Data Doug H-P @datadhp.bsky.social · 27/07/2026

This is a quick read - it gets at part of what has been bugging me about AI doing advanced mathematics. As we continue down this path, the question of “what are we even doing here? Why do we care about this in the first place?” becomes increasingly relevant. The answer quickly delves into philosophy

0111

Adam Kucharski @adamjkucharski.bsky.social · 25/07/2026

Faced with probability, AI can become a predictable parrot... My latest post on AI judgement, fuzzy concepts of likelihood, and why language matters: kucharski.substack.com/p/faced-with...

kucharski.substack.com

Faced with probability, AI can become a predictable parrot

How language shifts judgement

0162

Adam Kucharski @adamjkucharski.bsky.social · 24/07/2026

Latest post, on what happened when AI models with knowledge up to Jan 2026 made predictions about current events – and how language can change these assessments: kucharski.substack.com/p/faced-with...

063

Adam Kucharski @adamjkucharski.bsky.social · 22/07/2026

When researchers claim an AI model can now do end-to-end PhD quality research in a single prompt, I feel it reveals more about them and their idea of research innovation than it does about the model.

1664

Adam Kucharski @adamjkucharski.bsky.social · 20/07/2026

GPT is a pro at excuse euphemisms… “Did you actually look at the document?”

5759

Adam Kucharski @adamjkucharski.bsky.social · 20/07/2026

“I got Claude Fable to do this analysis and write this paper. I’m posting it here, but I haven’t checked it yet.” WELL CHECK IT THEN. WHY ARE YOU POSTING THINGS YOU CAN’T BE BOTHERED TO READ.

47812

Adam Kucharski @adamjkucharski.bsky.social · 19/07/2026

Some new additions to the collection this weekend… If you’re interested in uncertainty, evidence, AI and more and haven’t got a copy yet, you might want to head over to proof.kucharski.io

252

Reposted by Adam Kucharski

Uncertainty Principle @luckypetie.medsky.social · 17/07/2026

Excellent point about a possible inherent error in LLMs

094

Reposted by Adam Kucharski

Michael Plank @michaelplanknz.bsky.social · 17/07/2026

"LLMs are basically a turbocharged version of these uncertainty-inducing processing steps"

183

Adam Kucharski @adamjkucharski.bsky.social · 17/07/2026

New post, on the hidden uncertainties in many papers that use LLM judgement: kucharski.substack.com/p/when-it-co...

0297