AI unlocked effortless creation.
Effortless creation bred endless distraction.
Endless distraction left the field wide open to your own discipline.
Discipline always wins.
Effortless creation bred endless distraction.
Endless distraction left the field wide open to your own discipline.
Discipline always wins.
0
0
5
2,564
AI gave everyone the power to build, but robbed everyone of the focus to finish.
It gave tools to a distracted world, and a distracted world surrendered to its tools.
It doesn’t matter whether you are battling against the crowd outside—you only have to win the war against the clutter within your own mind.
It gave tools to a distracted world, and a distracted world surrendered to its tools.
It doesn’t matter whether you are battling against the crowd outside—you only have to win the war against the clutter within your own mind.
0
1
12
2,516
Getting answers is cheaper than ever, explanations are free to access for anyone with the Internet, but human thought. Human thought remains dear in every sense.
0
1
10
2,514
Everyone talks about AI replacing jobs, but few talk about how to use it for rapid upskilling.
AI won't learn for you, but it will break down any subject until you get it.
What we lack is not explanations, but the appetite for cognitive labour.
AI won't learn for you, but it will break down any subject until you get it.
What we lack is not explanations, but the appetite for cognitive labour.
3
3
12
2,924
I've been trying this out for a while now with an AI assessing how well I review my own papers and this study was using models that are considered ancient by AI standards.
Fable (even Opus 5.5) and Astra (even Sol 6.1) on high effort or above are unbeatable for me in their review quality and always find way more issues than I do.
As paper submissions are surging in most fields, it's time to rethink if we shouldn't be using AI for the review process.
Fable (even Opus 5.5) and Astra (even Sol 6.1) on high effort or above are unbeatable for me in their review quality and always find way more issues than I do.
As paper submissions are surging in most fields, it's time to rethink if we shouldn't be using AI for the review process.
Massive. New Harvard study finds that AI-written paper reviews now beat the best human peer reviews on every measure the authors tested.
Peer review is slow and often misses errors. Earlier studies concluded AI should only assist human reviewers, not replace them.
They used 134 a Harvard service that pays expert reviewers well. Claude Opus 4.8 reviewed the same 1124 papers, and GPT-5.5 plus 3 former journal editors judged the pairs blind.
GPT-5.5 picked the AI review in all 132 decided pairs, and the editors picked it 87% of the time. Only 5.5% of the AI's claims about the paper were wrong, versus 15.9% for the human reviews.
Teams that depend on expert critique can now let AI do the first full review pass, and keep a human for the final decision.
13
12
96
10,828
You can accomplish more meaningful research in 90 uninterrupted minutes than in a full day of fragmented effort.
Yes, the gap really is that wide.
Yes, the gap really is that wide.
2
9
56
3,879
Buried proof buys nothing.
153 experts have already taken this quiz. 80% share one blocker.
The blocker most people name first is socially acceptable. The real one is actually fixable once you see it.
See yours in 5 minutes: tally.so/r/1AEQy1
153 experts have already taken this quiz. 80% share one blocker.
The blocker most people name first is socially acceptable. The real one is actually fixable once you see it.
See yours in 5 minutes: tally.so/r/1AEQy1
0
0
0
246
Marks of a mature professor:
• Put student mentorship ahead of admin bureaucracy
• Letting minor deadlines pass without sacrificing sleep
• Decoupling your personal worth from citation metrics
The purpose was never to outwork the system.
• Put student mentorship ahead of admin bureaucracy
• Letting minor deadlines pass without sacrificing sleep
• Decoupling your personal worth from citation metrics
The purpose was never to outwork the system.
0
1
24
3,024
The biggest AI agent hack I know:
Pretend every task has a speed limit.
Discovery runs fast.
Private data crawls slowly.
Match your limits to the blast radius.
The autonomy that enables a breakthrough in enzyme research can just as easily escort an AI agent into a secure Medicare portal.
If you remember one thing before you set an agent loose, make it this.
Pretend every task has a speed limit.
Discovery runs fast.
Private data crawls slowly.
Match your limits to the blast radius.
The autonomy that enables a breakthrough in enzyme research can just as easily escort an AI agent into a secure Medicare portal.
If you remember one thing before you set an agent loose, make it this.
0
1
2
2,430
Uncertainty can be fun:
In a mystery novel.
In live sports.
In gambling.
AI can make serious decisions feel like harmless conversations.
It reassures you calmly.
It’s tempting to stop checking.
But your stakes haven’t changed.
Stay sceptical.
In a mystery novel.
In live sports.
In gambling.
AI can make serious decisions feel like harmless conversations.
It reassures you calmly.
It’s tempting to stop checking.
But your stakes haven’t changed.
Stay sceptical.
0
0
5
2,769
A longer AI answer can make you more confident in a mistake.
Longer explanations raise confidence.
They don’t help people catch wrong answers.
You read more.
You feel surer.
But you still need to check its assertion.
Read the research: nature.com/articles/s4225…
Longer explanations raise confidence.
They don’t help people catch wrong answers.
You read more.
You feel surer.
But you still need to check its assertion.
Read the research: nature.com/articles/s4225…
0
2
11
2,842
A stranger skims your 300 papers in one word:
Teacher.
WTF.
You're holding office hours.
You're rerunning the methods with a student.
You're 15 years into the claim they hear in 5 seconds.
Conduct the research they skipped over.
Win the decade they booked as teaching.
Teacher.
WTF.
You're holding office hours.
You're rerunning the methods with a student.
You're 15 years into the claim they hear in 5 seconds.
Conduct the research they skipped over.
Win the decade they booked as teaching.
0
0
7
2,884
How to turn any published paper into a writing template:
10 minutes.
~600 words.
1 highlighter.
1. Find a best paper in your field.
2. Look for 1 sentence that describes:
• Context (C)
• Problem (P)
• Method (M)
• Result (R)
• Contribution (K)
3. Get the skeleton and the style.
4. Write it in your own words.
Reusable and good.
10 minutes.
~600 words.
1 highlighter.
1. Find a best paper in your field.
2. Look for 1 sentence that describes:
• Context (C)
• Problem (P)
• Method (M)
• Result (R)
• Contribution (K)
3. Get the skeleton and the style.
4. Write it in your own words.
Reusable and good.
2
10
69
6,597
Buried proof buys nothing.
153 experts have already taken this quiz. 80% share one blocker.
The blocker most people name first is socially acceptable. The real one is actually fixable once you see it.
See yours in 5 minutes: tally.so/r/1AEQy1
153 experts have already taken this quiz. 80% share one blocker.
The blocker most people name first is socially acceptable. The real one is actually fixable once you see it.
See yours in 5 minutes: tally.so/r/1AEQy1
0
0
0
368
We chase certainty to avoid regretting our decisions.
Turn to political promises.
Read another paper.
Ask AI.
Eventually, you must choose based on what you know.
It's uneasy.
It's awkward.
It's uncomfortable.
Behind the painful distress lies satisfaction.
Don't build your identity around having the right answer.
Build it around the process of choosing one.
Turn to political promises.
Read another paper.
Ask AI.
Eventually, you must choose based on what you know.
It's uneasy.
It's awkward.
It's uncomfortable.
Behind the painful distress lies satisfaction.
Don't build your identity around having the right answer.
Build it around the process of choosing one.
0
0
8
2,739
Instead of watching one hour of Netflix tonight, watch… NO! Just take the time off and watch/read/do whatever you want. Your work is hard enough without influencers telling you what dumb video will make you more intelligent. Enjoy your well-deserved break.
1
1
14
2,881
Content mass production is killing creativity.
AI has made us excellent at producing content.
Chattie just lurves to help out.
And we gobbled it up.
The thing is:
You don't understand the sh*t you didn't do.
You cared about the artifact.
You ignored the process.
Two ways to fix this:
AI has made us excellent at producing content.
Chattie just lurves to help out.
And we gobbled it up.
The thing is:
You don't understand the sh*t you didn't do.
You cared about the artifact.
You ignored the process.
Two ways to fix this:
1
0
8
2,879
How to build a stellar AI-supported system:
1. Understand your start (A) and end points (B)
2. Capture the process of how you get from A to B
3. Review and examine your capture for improvements
4. Build checklists with AI for repeatable tasks
5. Structure where you save reference information
6. As you redo the process improve reference info
7. Redo only the parts of the process valuable to you
8. If AI does the parts not valuable to you → Review output
9. Review the quality of your process output
This is how you can build valuable systems.
Understand that some non-pleasant parts still teach you.
Never trade in learning opportunities for automation.
1. Understand your start (A) and end points (B)
2. Capture the process of how you get from A to B
3. Review and examine your capture for improvements
4. Build checklists with AI for repeatable tasks
5. Structure where you save reference information
6. As you redo the process improve reference info
7. Redo only the parts of the process valuable to you
8. If AI does the parts not valuable to you → Review output
9. Review the quality of your process output
This is how you can build valuable systems.
Understand that some non-pleasant parts still teach you.
Never trade in learning opportunities for automation.
1
0
2
393