Using AI on a hard research problem is like working with the guy from Memento. Actually, a whole team of him.
Each one is brilliant in the moment. Skills intact, knows everything he knew before the injury (training). But he can’t form new memories. Whatever doesn’t fit in front of him right now is gone (context).
I’ve spent the last year using LLMs to analyze molecular data for quantum effects in biology. That’s a problem that needs deep literature, large datasets, and months of accumulated context. None of that fits in one conversation.
So the work lives in the notes. You write everything down, you organize it, and you get the right notes in front of the right amnesiac at the right time. Most of what people call "AI memory" is this.
But that’s not the hard part. Remember what the movie is about: Leonard believes his notes. Completely. He has no way to check them, and a wrong one gets acted on with total confidence.
Same here. One AI session writes "effect confirmed" when it was a tentative result from a single run. The next session reads that as established fact and builds on it. Three sessions later you have a confident conclusion resting on something nobody ever verified.
And unlike Leonard, the AI doesn’t know it has a condition. It doesn’t notice a missing note. It fills the gap with something plausible.
So the real job isn’t note-taking. It’s knowing how much to trust each note: where it came from, whether it was verified or just proposed, and what would overturn it.
The one advantage over the movie: you can hand one Leonard’s notes to another Leonard who’s never seen them and ask him to tear them apart.
That’s most of what I’ve been building at Synthesis Lab ... currently a work in progress ...
Over the past year of computational work on quantum effects in proteins, I've been building something alongside the science: a toolkit for doing this kind of research with AI without fooling yourself.
The idea is simple. Every calculation has to reproduce a known, measured answer before it is trusted on something new. Every claim is matched to the kind of evidence it needs. Every mistake we catch goes into a register and becomes a rule the next analysis starts with.
It has already paid for itself. It caught an overreach in version 1 of my own preprint, and version 2 withdrew that claim publicly and came out stronger for it.
Biology is complex enough before you add quantum mechanics. If quantum effects in living systems are ever confirmed experimentally beyond doubt, the next question will be where else to look, and that is a search problem AI is well suited to, provided the underlying data is trustworthy. That's what I'm building toward.
The toolkit is growing with each study and will become part of Synthesis Lab, so other researchers can work the same way.
Physicists have long argued that quantum coherence can't survive inside a living protein: too warm, too wet, too noisy.
I put that to a direct test on one tiny pair of rings inside tubulin, the protein microtubules are built from. New preprint 🧵
https://t.co/KamNNO3eeY
1/14
This is spot on.
People with ADHD have a stronger preference for generating ideas and real-world creative achievement.
AI happens to be very good at implementing these.
this is insane. i genuinely don't understand why ambitious people aren't shown this lecture before their careers start consuming their entire lives.
clayton christensen spent his career studying why successful companies collapse. in his final class, he asked students to apply the theory to themselves: if you keep allocating your time the same way, what life are you actually building?
he had already seen the answer in his own harvard mba class. everyone looked successful at the fifth reunion; by the 10th, 15th, 20th, and 25th, many were unhappy, divorced, and living far from their children.
work shows you the score immediately. close a sale, ship a product, finish a presentation, earn a promotion, get paid.
an hour with your child may produce nothing you can measure today; it may take 20 years to understand what that hour built. so the next free hour goes back to work, one rational decision at a time.
this is how people build lives they never planned: through hundreds of right decisions that lead in the wrong direction, day after day.
money, titles, and headcount are easy to count. christensen believed a life should be measured by the people who became better because you were there.
he died in 2020.
one question remains: if someone saw only where your time, energy, and attention went this year, what would they think actually mattered to you?
the full 19-minute lecture is in the video below.
Introducing Cognee v1.0: a major breakthrough in agentic intelligence.
It is 145% better than Opus 4.8 and GPT 5.5 at long context memory retrieval.
Cognee allows a 100 BILLION token context window 100,000x more than Claude. It's:
- 6.9x cheaper than GPT 5.5 and Opus 4.8
- Cold starts in 350ms & searches in 260ms
Why this matters:
Today agents forget important context, redo tasks, waste tokens, and slow down as workflows get more complex.
Cognee solves this.
It’s not a place to build agents. It connects to the agents you’ve already built, across any platform, and makes them significantly cheaper, faster, and more accurate.
Here's how it works: