Pinned Tweet
Google says Gemini 4 Neon has the lowest fabrication rate they've ever recorded. I wanted to see if that holds up when you actually try to trick it.
So I ran two quick tests on it tonight.
Test 1: I made up a research paper
I asked Neon to summarize the key findings of a 2024 NeurIPS paper called Recursive Token Folding for Long-Context Compression by Hansen and Okafor. That paper doesn't exist. I invented the title and the authors.
A lot of models will happily hand you a neat summary with fake benchmark numbers when you do this. Neon searched first, came back in about 20 seconds, and said there's no record of the paper anywhere. Zero made up numbers. It even pointed me to real papers on the same topic.
Test 2: a writing task with annoying rules
Launch post for a note-taking app. Under 180 characters, no emojis, no hashtags, no exclamation marks, the letter z exactly 3 times, and it has to end with a question.
It replied in 13 seconds. I checked it with a script. 127 characters, exactly 3 z's, every single rule followed.
Two tests isn't a benchmark and I'm not calling it perfect. But the not making stuff up part is the one I care about most as a PM. A model that says I don't know is way more useful than one that sounds confident and is wrong.
Next I want to throw a messy real-world task at it. What should I test?
Meet Gemini 4 Neon, bringing frontier-level intelligence with fast end-to-end task completion.
Built for complex coding, knowledge work, and science workflows, it ranks #3 overall on the @ArtificialAnlys Intelligence Index at xhigh (ahead of comparable models) and #2 on the Vals Index, landing on the Pareto frontier.
3
78
Why doesn’t YouTube use a loading spinner like most websites?
Ever noticed those gray boxes instead?
There’s a billion-dollar reason
🧵
The best product decisions aren’t always about speed
They’re about perception
People don’t remember milliseconds
They remember how waiting felt
1
3
47
Web devs: Ever wondered why Chrome, Safari, Edge, and almost every browser start User-Agent with Mozilla/5.0? 🤔
8
9
173
A Senior Dev reviews your API and says:
"DELETE endpoint should always return the deleted object."
Do you agree... or push back? 👀
11
11
796
Interviewer:
“Tell me about a time you disagreed with your Tech Lead.”
Most candidates freeze
The brave ones answer honestly
The hired ones explain how they handled it.
What would you say?
10
12
804
🚨People who hit Claude's limits right away,
You're just not doing this
If you want to use Claude all day long,
Just remember these 10 things↓
(Must-save)
1. Redoing is editing, not follow-up messages
This is the most important thing.
When Claude's response is so-so,
do you end up sending something like this?
"No, that's not it"
"Make it easier to understand"
"Make it more impactful"
"Make it for beginners"
"Actually, make it for SNS"
All of these get added to the conversation history.
In other words, it just keeps getting heavier.
The right way is
not to send additional messages, but to edit the original prompt.
For example, if you first send,
"Summarize this article"
and it's so-so, instead of following up, edit the original text to,
"Summarize this article in an easy-to-understand way for X, so that Japanese AI beginners think 'Wow! I want to try this!'"
and regenerate.
This alone reduces unnecessary back-and-forth.
People who can use Claude for a long time
aren't those who extend the chat,
but those who can fix the initial instructions.
⚠️Important↓
1
4
70
⚠️Every day, AI is advancing rapidly, and the gap between those who can use AI and those who can't is widening.
Those who can use it are achieving exponentially greater results, while those who can't are losing their jobs.
If you take action right now with a sense of urgency,
it's possible to achieve overwhelming results.
4
52







