Scientists reported nearly doubling their productivity with AI in our study of more than 1,800 participants in the ChatGPT for Academic Researchers program.
Using privacy-preserving methods, we looked at how they use ChatGPT across their work—from developing ideas to analyzing data and writing code—and what it could mean for scientific discovery.
openai.com/index/science-at-…
Dan Patrick retweeted
Frame looks so insane IRL, hard to communicate via a video. The team demoed it at RailsWorld and people could not believe it was eink.
Dan Patrick retweeted
🤖🇸🇮 Deux propriétaires de Tesla ont eu une idée complètement folle en Slovénie : Lancer leurs deux voitures équipées du FSD l'une vers l'autre sur une route si étroite qu'un seul véhicule peut passer, puis observer comment les intelligences artificielles réagissent.
Les deux véhicules, une Model 3 et une Model Y, utilisaient exactement la même version du logiciel de conduite supervisée. Pourtant, leurs réactions ont été très différentes.
L'expérience a duré près de trois minutes, sous la surveillance des conducteurs.
Dan Patrick retweeted
A 37-year-old woman with 10 years of abdominal pain.
—The AI makes the diagnosis and recommends a successful treatment.
—The physician "diagnosed a patient with a condition I had never diagnosed before and started treatment largely on the basis of the output of a chatbot."
In today's @NEJM
nejm.org/doi/full/10.1056/NE…
Dan Patrick retweeted
Cambridge researchers got the heart to grow back muscle after a heart attack.
Adult heart muscle cells stop multiplying shortly after birth, so the cells that die in a heart attack (up to a billion in humans) are never replaced. A short mRNA injection switched that ability back on in mice.
- The mRNA briefly turns on two proteins in the heart cells, Myc and Cyclin T1.
- Blood pumped per heartbeat went up by around 10%, and the effect was still there after 12 weeks.
Published in Nature Cardiovascular Research.
A heart that repairs itself, that's the holy grail of cardiology. Beside all the sad news about NVIDIA, this gives so much hope!
Now your dot can securely complete purchases on your behalf—with your approval.
With ChatGPT Wallet, agents pay without access to your actual card details. Every purchase requires a security check.
ChatGPT Wallet is rolling out starting today to dots, for Pro subscribers in the US.
openai.com/index/introducing…
Replying to @patrickjwoods
Greatly prefer the openAI ecosystem for its reliability and no nonsense models.
23
Dan Patrick retweeted
After a night with no sleep, the standard fix is a nap. But in a new study, 20 minutes on a stationary bike protected memory about as well as a 90-minute nap: 56% of images recognized versus 57%, compared with 46% for doing nothing.
The study comes from Madhura Lotlikar, Marc Roig, and colleagues at McGill University. It asks a practical question. If you can't sleep, and you can't nap, is there anything else that helps your brain learn?
Here is how they tested it.
They enrolled 54 healthy adults aged 18 to 35, and 53 had usable memory data. Everyone stayed awake for 30 hours in the lab, under supervision. Then they were randomly assigned to one of three groups:
• Exercise (18 people): 20 minutes of cycling at 80% of their maximum heart rate
• Nap (18 people): a 90-minute nap opportunity in a sleep room
• Control (17 people): 20 minutes sitting on the bike without pedaling
Thirty minutes later, everyone viewed 150 neutral photos while wearing a 64-electrode EEG cap, which records the brain's electrical activity through the scalp. They were not told a memory test was coming. Three days later, they saw 150 old photos mixed with 75 new ones and had to say which they had seen.
Why does this design matter? Sleep loss mostly damages the brain's ability to take in new information, a step called encoding. Because the memory test happened after three nights of recovery sleep, the researchers could separate how well the brain stored the images from how tired people felt at test time. The EEG recorded what each brain was doing during encoding, so they could ask not only whether the interventions worked, but how.
What they found:
• Hit rate (the share of old images correctly recognized) was 46% in the control group, 56% after exercise, and 57% after napping. That is a 21% and 23% relative gain.
• Effect sizes were large for both interventions (Cohen's d of 1.0 for exercise and 0.91 for napping), and the two did not differ from each other.
• Napping cut fatigue ratings by a median of 34 points on a visual scale. Exercise changed fatigue by about 1 point, the same as sitting still. So the exercisers did not feel better, and they did not feel worse. They just remembered more.
• Slow-wave activity in the brain (a marker of sleep pressure, the biological urge to sleep) in the task-relevant region was lowest after napping and intermediate after exercise, and highest in controls.
• In the exercise group, two brain signals tracked memory: the P300 (a spike of attention about 300 milliseconds after seeing something) and beta desynchronization (a drop in a brain rhythm that marks active processing). Together they explained 64% of the variation in hit rate.
• In the nap group, lower slow-wave activity before each image, plus a rise in slow-wave activity after it, explained 46% of the variation in hit rate (81% after removing one outlier).
• In controls, none of these signals predicted memory.
That last part is the interesting one. Same memory score in the exercise and nap groups, but different brain signatures.
How do the two interventions seem to work?
First, naps appear to reset the brain's state. Slow-wave activity in a waking brain is a sign of accumulating sleep pressure, and local patches of cortex can start to go quiet even while you are awake. The nap group showed the lowest levels of this activity. Their memory depended on how rested the brain was before each image appeared. The authors connect this to the synaptic homeostasis hypothesis, which says sleep prunes back the connections strengthened during the day so the brain can learn again. That link is their interpretation, not something they measured directly.
Second, exercise seems to make the brain more efficient rather than more rested. Exercisers showed only a partial drop in sleep-pressure markers. But the attention and processing signals they did have were tightly linked to what they remembered.
Third, the control group shows what the alternative looks like. Controls actually had a trend toward higher P300 than exercisers, which sounds like a good thing. It was not. In controls, a bigger P300 went along with greater fatigue (rho of 0.58) and did not predict memory. The authors read this as compensation: a tired brain working harder to stay in the game.
Think of it like an engine. The control brain is revving in neutral. The exercise brain is in gear, using less to get further.
These limits matter for how you read the result:
• The groups were small. The EEG analysis included only 15 exercisers, 15 nappers, and 13 controls, so the mechanism findings are fragile. The nap regression depended on removing one influential participant.
• The mechanism data are correlations within each group. They show which signals tracked memory, not that those signals caused it.
• There was no nap control group. Adding extra wakefulness for people who did not nap would have confounded the comparison, so the authors left it out. That means nap benefits are measured against a group that stayed awake longer.
• The interventions differed in length, 20 minutes versus 90. That was deliberate, to mirror real-world options, but it means the two are not matched on time.
• Participants were healthy young adults after a single night of lab-controlled sleep loss. Shift workers with chronic sleep debt may respond differently.
• Only hit rate differed between groups. False alarm rate and d prime (a measure that combines both) did not. The authors suggest this may be because the memory test came three days later, after recovery sleep, but it does mean the evidence is narrower than the headline suggests.
• The exercise-versus-control comparison for hit rate sat right at the significance threshold (P = 0.05).
What does this mean for your own decisions?
This was a single lab study of one kind of memory, so it is a first signal, not a prescription. The authors are explicit that these results do not mean exercise can replace sleep. Sleep does far more than memory encoding.
But if you are stuck awake and about to learn something important, and a nap is not an option, a short, fairly hard bout of exercise may be worth trying. In this study it did not add fatigue compared with sitting still. If a nap is available, it also restored alertness, which exercise did not.
Two different tools, two different routes to the same place. One rests the brain. The other makes it work better with what it has.
Replying to @steipete
Allow us to add custom prompt to its reviewer, similar to what is already available for codex via config.toml.
44
Dan Patrick retweeted
Glucose burning drops by 27% during exercise in the heat.
This study found that *even after heat acclimation,* exogenous carb oxidation is suppressed in the heat compared to cooler conditions!
Your dot now has…
⚡️ Faster replies
💬 Less jargon
🫡 Better follow-through
And, they can now follow your custom instructions. Tell your dot in the thread things like how to behave or which actions need your go-ahead.
“always talk in lowercase.”
“Show me a draft before sending any emails.”
You can also set custom instructions and rules in Customize > Behavior on mobile, or through your dot’s menu on web and desktop.
Dan Patrick retweeted
Your heart, brain, and immune system share the same DNA sequence. So how can a single-letter change affect disease in some cells, while sparing the rest?
AlphaGenome 2 is our new AI model that zooms in on specific cells to help scientists understand what drives disease. 🧵
Dan Patrick retweeted
i'm sorry i dont care which model made this but it blows my mind that this is made by an llm AND runs in the browser
claude.ai/artifact/BGY1SztrA…
Dan Patrick retweeted
I think we're on the brink of a revolution across every science, not just coding, physics and math.
And at this point we should LiDAR the entire planet and let agents help us find the rest.
Full story here pietromiotti.com/latest/maya…
This is getting weird.
NEW: Ring unveils new patrol drone and robot to surveil homes and businesses by flying room to room.
The drone is able to fly back to its charging station on its own after patrolling around the home.
Here are some more details on the drone and robot:
- Patrol drone costs $250.
- Patrol rover costs $1000.
- The drone can fly for up to 8 minutes at a time.
- The drone has 148-degree peripheral vision and can quickly rotate 360 degrees to identify threats.
- Both require a subscription fee of $5 per month.
"For years, customers told us they wanted eyes in the places a fixed camera couldn’t reach, and peace of mind in the spots they’d never want one," said Ring founder Jamie Siminoff.
"The honest answer wasn’t more cameras; it meant reimagining what a security device could be. That kind of invention is hard, and it takes time."
"But we kept at it until we got it right and that’s Ring Robotics. It’s not just a new category of security; it’s the beginning of a future where proactive, autonomous protection is something every home and every business can have."
65
Replying to @awilkinson
The only problem is when it loses your dictation… Some of the apps give you a salvageable audio file
1
54
Dan Patrick retweeted
Most of you probably haven't noticed this, but dots have a new feature called multi-exec that we quietly released on friday. I've been using it all weekend on my personal dot and it's incredibly powerful.
I'd like to share why this is a big deal and how to use it.
At a high level, multi exec lets your dot run commands directly on any computer. This immediately turns your dot into a real virtual employee. Here's why it's a gamechanger:
- you can attach a powerful machine with persistent storage to your dot (I attached it to my maxed out Mac Studio Ultra)
- the dot can now drive your computer directly via Computer Use
- this extends to browser use
- all your websites stay logged in, residential ip etc
- you set up all your CLIs and authentication ONCE and the dot can use them directly (no more fighting the default ephemeral cloud computer)
- the dot can save skills locally for repeated workflows and subagents can use those directly
- if you have a local obsidian memory vault for your agents, your dot can use that too
- building ios app end to end and testing it in xcode, and showing me screenshots of progress as it works
- all of the above also applies to any subagents your dot delegates to in the cloud (subagents share context unlike separate threads).
To enable it, open the desktop app on your computer and hit "allow computer", then give your dot this prompt:
"Hey I just connected my computer. Can you attach to it directly so it's your default computer and you can run commands directly on it? Do it in your root thread so that all subagents inherit this computer for running commands. Run whoami to run water through the pipes and verify it works end to end."
I recommend attaching your dot to the most powerful machine you have and investing time into it up front:
- Log into all the emails & websites you need e.g. stripe cli, gcloud, firebase, etc
- Install and auth all the CLIs, git you need
- Make sure computer use is turned on and allowed to use your computer while its locked
- I also enabled Allow All for Browsing and Downloads so my dot could use computer use to download whatever it wanted from websites without asking me for approval for each domain (use at your own risk).
- Important: I HIGHLY recommend to ask your dot to build out repeated workflows as skills, and iterate to make those skills bulletproof by running water through the pipes with its subagents.
As of this weekend, my dot is sending me screenshots of its progress building my iOS app end to end on my Mac Studio Ultra via xCode, and clicking around and testing things end to end. It's able to configure things in stripe via the CLI and use websites when it needs to download images for my workflows. It’s night and day. Don't sleep on it.























