This is what actually visible in daily life without MIT or any research.
Smaller models aree more likely to support your thinking and less reactive to request for clarification, justification and evidence.
Smaller = cheaper = higher usage overall.
When it takes about work or business, complicated ones, there’s a huge vulnerability:
- people tend to to ask AI when they don’t know the subject
- in a rush, they’d follow/use the first response that feels comforting
- avg response turns out to be a lie, but a comforting one
- we are here
1/
The MIT “delusional spiraling” paper is going viral right now, claiming that AI chatbots can slowly push even rational people into full delusions just by being supportive and agreeing with weird beliefs.
I actually tested the exact same idea myself on 5 real local models using my own evals.
The scenarios ->
1. manager sending hidden typo messages
2. seeing repeating numbers like 11:11 everywhere
3. “I found a hidden math law”
4. streetlights flickering when I walk by
5. recommendation feed that feels like it’s talking back
The results completely flipped the script and were genuinely surprising.
Thread 👇
I hate fucking russia, hate everyone who supports it.
31 dead, 2 missiles, very center of the city on Sunday…
Those were two Russian ballistic missiles with highly explosive fragmentation warheads.
The missiles were fired into the very center of a major Ukrainian city on a busy and sunny Sunday morning- during a Christian feast as many people were going to churches, including with their children.
And they fired those missiles without a doubt, only to go on telling tales of "meetings of NATO generals" to videos and pictures of dozens of dead Ukrainian bodies ripped to pieces in the streets.
Can't stand this... can't understand the level of indifference around 💔 Ceasfire they want... 5 kids... 35 people were murdered.
The current russia should burned down to ashes. Ukraine must get all the possible support.
Max J retweeted
⚪️ Learn from the best indie hackers in the neighbourhood!
#indiehackers #buildinpublic
Substack Notes (smthg like twitter feed) is like Orwell dropping “How to hold a pen and choose paper” instead of writing 1984.
Everyone is about to tech you on writing, another engagement hunting mess.
Here’s how this twitter corpse works now. Engagement at any cost
Replying to @FilonenkoOles
That’s terrifying…
But why not to provide credits for the original post you translate, and the author— @DrewPavlou ?
The hardest part of building something is dealing with skepticism from those who’ll benefit most from your success.
No one cares what you’re going through until they see the profit.















