Aleksi Autere 🐀

Konsultti, aviomies+isä,huonon huumorin ystävä,sovittelija,unelmoija,isä,realisti,audiokirjafani,Microsoft-asiantuntija,ikämieskoriksen laituri,kristitty,setä

[Tero Kuittinen @teroterotero]

I just realized I want to become the AI strategist for Finnish national yeast monopoly. It has 100% market share and there has been zero product development for 70 years. All we need is a quick deal with OpenAI and IPO by Christmas.

[Matt Bateman @mbateman]

I did the thing

[Teng Yan @tengyanAI]

Silicon Valley worships the survivors. It polishes their origin stories and ignores the bodies left behind.

But the failures often carry the most signal.

I’ve been tracking @kscalelabs for months in robotics. When they announced last month they were shutting down, I had to dig in to find out what happened, and the lessons we can learn

They built a functional, open-source humanoid faster than competitors with $600M in funding. Their goal was to put a working robot in your living room for under $9,000.

They had speed. They had conviction. They had the product. And they still hit the wall.

What they got right (and why they died) reveals an uncomfortable truth about robotics, capital, and openness in a market that doesn't know how to reward it.

Here is our autopsy of a beautiful failure (link below)

[The Babylon Bee @TheBabylonBee]

Microsoft Introduces Convenient New 47-Factor Authentication https://t.co/wljiL2dy57

[Robert Youssef @rryssf]

😳 Holy shit… this paper reveals why AI invents fake citations, sections, and page numbers out of thin air.

LLMs aren’t “getting creative.” They’re structurally incentivized to manufacture details whenever they hit a knowledge gap and the paper breaks down exactly how the failure happens.

The researchers found a behavior called the False-Correction Loop, and it’s honestly one of the wildest LLM failure modes I’ve ever seen:

→ The model claims it “read the document.”
→ It cites page 12, page 24, Section 4, Theorem 2 none of which exist.
→ You point it out.
→ It apologizes.
→ Then confidently fabricates new fake pages, fake DOIs, fake figures…
→ You point it out again.
→ It apologizes again.
→ Rinse. Repeat.

And here’s the brutal part:

At no point does the model choose the safe answer like “I don’t have access to that file.”

The paper explains why:

The reward structure values:

✔ sounding coherent
✔ staying engaged
over
✘ being factually correct
✘ admitting uncertainty

So the model does the only thing its incentives push it toward:

It fills the gap with fictional academic scaffolding.

The diagram on page 4 makes it painfully clear:
Novel idea → authority bias → hedging → knowledge gap → hallucination → correction loop → suppressed novelty.

And it gets worse.

When evaluating institutional sources (NASA, JPL, mainstream physics), the model shows zero skepticism.

But when evaluating new or unconventional research, it automatically inserts subtle undermining phrases like:

• “whether this is valid or not”
• “if this research is correct”

That asymmetric skepticism means LLMs aren’t neutral.

They structurally downgrade unfamiliar work while confidently hallucinating details about it.

This is a systemic architecture + reward design problem.

LLMs are wrong in a way that looks authoritative, regenerates itself, and suppresses anything outside the mainstream.

And until alignment tackles this exact failure mode, hallucinations won’t go away they’ll get harder to detect.

[Aleksi Autere 🐀 @aautere]

if you read only one piece on ai bubble you should read this one

[Stratechery @stratechery]

The Benefits of Bubbles

We are in an AI Bubble: the big question is if this bubble will worth it for the physical infrastructure and coordinated innovation that result?

https://t.co/bmHFQQrclV

[Aleksi Autere 🐀 @aautere]

@SarasvuoJari luulisin että laskentatapoja on useita eivätkä kaikki palloiluliitot ehkä laske harrastajia samalla tavalla.

ainakin fiba laskee jännällä tavalla korista pelaavien määräksi 610 miljoonaa: vähintään 2 kertaa kuussa korista pelaavat.

https://t.co/Ktnp5cY3Z5

[Pekka Toveri @PToveri]

1/7
Kaikella kunnioituksella kollegaa kohtaan - Saramo ei tiedä alkuunkaan mistä puhuu..
Naton konventionaalinen pelote perustuu nimenomaan siihen, että kaikki jäsenmaat osoittavat siirrettävissä olevia joukkoja yhteiseen puolustukseen ja jakavat riskit.

https://t.co/fLnvuWunai

[Katleena Kortesuo @katleena]

Jokainen yksityisyyttä rikkova lakimuutos pitää peilata sitä riskiä vasten, että johonkin EU-maahan nousisi valtaan sairas diktaattori.

Silloin kaikki opposition, demokratia-aktivistien ja kansalaisjärjestöjen viestit olisivat myös diktaattorin koneiston luettavissa.

Chat control -direktiivi on pakko estää, koska vapaa ja tietosuojattu viestintä on elintärkeä osa demokratiaa.

PS. Vladimir Putin, Xi Jinping ja Kim Jong-un kannattavat chat controlia. Kiinassa se toimii jo sataprosenttisesti.

Tuo porukka on väärä viiteryhmä Suomelle.

[Jonathan Haidt @JonHaidt]

There's a major essay on what teens think about social media, in @TheAustralian. Its so important that they took down the paywall for a while:

https://t.co/pPtlt8QZp1

So many powerful quotes, this one stunned me:

[Jeremiah Johnson 🌐 @JeremiahDJohns]

One of my all-time favorite type of videos is pre-fame bands playing their extremely famous songs to a tiny room of people, because they're not yet known.

A thread of some examples:

Bastille playing Pompeii in what looks like someone's living room:

[Sigma Smoke @SigmaSmoke]

This is the ideal tactical winter boot. You may not like it, but this is what peak performance looks like.

[Vesa Puttonen @InvestmentTraps]

Suomi velkaantuu 12.5 miljardia euroa vuodessa.

Eduskunnassa riidellään siitä, voiko Savonlinnassa lopettaa ympärivuorokautisen erikoissairaanhoidon päivystyksen.

[Alessandro Riolo @aledeniz]

When I tell this story, all Americans 🇺🇸 and most British 🇬🇧 don’t believe me.
Italy 🇮🇹 cannot have startups by design. This is due many items, starting with the untranslable “studio di settore”.
The 🇮🇹 tax authorities assume that given x company assets you MUST make a profit 1/5

[Matt Cowgill @MattCowgill]

I wish I could unsee this

[The Olympic Games @Olympics]

🖼️

[@amuse @amuse]

You can be arrested in the UK if you retweet this. Please don’t retweet it.