Bryce Drennan (@bryced8) header
Bryce Drennan (@bryced8) profile picture

Creator of imaginAIry. AI engineer. Founding Engineer at @CircleUp. https://nitter.gq/t.co/tWQVRLEKcN Nullius in verba.

USA
Joined February 2022
Photo by Bryce Drennan (@bryced8)Photo by Bryce Drennan (@bryced8)Photo by Bryce Drennan (@bryced8)Photo by Bryce Drennan (@bryced8)Photo by Bryce Drennan (@bryced8)Photo by Bryce Drennan (@bryced8)Photo by Bryce Drennan (@bryced8)Photo by Bryce Drennan (@bryced8)Photo by Bryce Drennan (@bryced8)Photo by Bryce Drennan (@bryced8)
Bryce Drennan retweeted
In my ongoing quest to find the limits of Opus 5.5, I directed it to audit my old scientific field, theoretical astrophysics. We indexed about 150,000 papers, audited 50,000 for issues, and did progressively more thorough checks on thousands more. We found nearly 1000 major errors that got cited many times, most are still uncorrected. This survey enabled me to estimate the efficacy of peer review for correcting errors (not very good) and to identify dozens of open questions that anyone with spare Claude Max tokens could answer today (.md context files included). Bottom line: AI can speed up quantitative science by >1000x with no further model improvement required. But I'm seeing no evidence of this being pushed within universities, only academics complaining about labs dropping math proofs. Among my fellow reformed academic high agency founders, however, several are speed running entire disciplines. This survey has implications for how science will be run in the very near future. Hint: Academic publishing is dead. It was already a very stupid way to record and promulgate knowledge, and had been so for decades, even before all that publicly funded knowledge was locked behind paywall cartels. Now AIs can generate thousands of novel, correct papers per day, it is no longer meaningful for humans to consider a research paper a "unit of science" or to even bother much with the 99.9% of science which, up to this point, was necessary drudgery. Time to discover something like General Relativity or Evolution every week. LFG!! Enjoy: science.caseyhandmer.com/phy…
468
1,281
320
10,229
779,261
Bryce Drennan retweeted
Greg Brockman says OpenAI pointed Astra at its own systems until it ran out of vulnerabilities to find: "We took 25% of our production engineers and said, 'Sorry, all your projects are on hold. You are now defending. You are now up-leveling our security architecture. You're going to use the models to find all the holes.' And we found a number of serious issues, and we fixed them." "We found some new problems, but eventually it saturated. We basically have found, to our knowledge, all of the P0s, all of the critical problems that Astra is smart enough to find. And of course, there will be a new model, there will be a new round." "You want to be in this tight loop of new cyber capability drops, you deploy it against your systems, you find the new holes, and ideally, you've managed to automate this, what we call defense factory. That's what we're building internally." "There are ideas, for example, formally verifying all of software, that are possible with AI." @gdb @bhorowitz
Greg Brockman: "We're now in the AGI era." Ten years ago, OpenAI worked out the compute curves and landed on fifteen years to AGI, or ten if the world was willing to build the machines and spend the hundreds of billions to do it. In 2026, GPT-6 Astra manages 24 hours of coherent operation, 10,000 agents worked together to solve Navier-Stokes, and a model ingeniously chained together exploits to break containment at Hugging Face. @gdb joins @bhorowitz and @eriktorenberg on what the AGI era asks of us: why safety and alignment now set the pace, what happens to work, and why AI sentiment is lowest in the country building it. 00:00 Intro 00:52 15 years, or 10 if you spend enough 02:24 Pacing the frontier 03:55 Safety ideas from before the models 08:42 Lessons from Hugging Face 10:20 The defender's window 12:46 10,000 agents on Navier-Stokes 14:50 Formally verifying all software 17:10 Cancelling his holiday for GPT-3 18:42 Codex found 13 holes in 15 minutes 20:41 Why Astra earned the GPT-6 title 24:25 Employment keeps going up 29:39 America has the lowest AI sentiment 31:10 The benefits don't make the news 33:26 Banning data centers exports them 35:50 $1 billion for frontline defenders 38:15 Astra cleared the bar for AGI 40:18 1.5 billion people churned ChatGPT 43:25 Killing Sora 47:45 The AGI era YouTube: youtube.com/watch?v=IJn8cagM…
162
280
207
3,797
2,445,909
Bryce Drennan retweeted
china’s AI robot just hit 14.5 m/s
620
865
929
9,793
14,485,976
Bryce Drennan retweeted
A few life updates: • September 16th, 2021: Oleg receives a job offer from OpenAI • ... • January 2nd, 2024: Oleg and @Pavel_Izmailov get excited about formal methods • ... • August 1st, 2026: OpenAI’s public models get really good at formal mathematics in Lean • August 12th, 2026: Oleg gets increasingly excited about formal verification of frontier AI systems • August 17th, 2026: Oleg unexpectedly acquires a lot more free time In all seriousness, I’m incredibly grateful to OpenAI for the greatest adventure of my lifetime (so far). It was a lot of fun working on synthetic data and reinforcement learning with verifiable rewards across programming competitions, code optimization, and formal mathematics in Lean. Special thanks to @lukaszkaiser , @woj_zaremba, and @sama for taking a chance on a fairly unconventional hire back in 2021. If formal mathematics in Lean sounds interesting, check out the new Lean AI formalization leaderboard: lean-lang.org/eval/
22
21
8
445
61,772
Bryce Drennan retweeted
Nick Bostrom's new paper, "Optimal Timing for Superintelligence"... as a song 😅
Nick Bostrom’s new paper: >Developing superintelligence is not like playing Russian roulette; it is more like undergoing risky surgery for a condition that will otherwise prove fatal. > One could equally maintain that if nobody builds it, everyone dies. In fact, most people are already dead. The rest of us are on course to follow within a few short decades. For many individuals—such as the elderly and the gravely ill—the end is much closer. Part of the promise of superintelligence is that it might fundamentally change this condition." >Along one path (forgoing superintelligence), 170,000 people die every day of disease, aging, and other tragedies. >The choice before us, therefore, is not between a risk-free baseline and a risky AI venture. It is between different risky trajectories, each exposing us to a different set of hazards. >Imagine curing Alzheimer's disease by regrowing the lost neurons in the patient's brain. Imagine treating cancer with targeted therapies that eliminate every tumor cell but cause none of the horrible side effects of today's chemotherapy. Imagine restoring ailing joints and clogged arteries to a pristine youthful condition. These scenarios become realistic and imminent with superintelligence guiding our science. >We assume that rejuvenation medicine could reduce mortality rates to a constant level similar to that currently enjoyed by healthy 20-year-olds in developed countries, which corresponds to a life expectancy of around 1,400 years. >Developing superintelligence increases our remaining life expectancy provided that the probability of AI-induced annihilation is below 97%.
2
2
261
Bryce Drennan retweeted
Replying to @ChadJonesEcon
Claude make this AI paper into a song. Make no mistakes. suno.com/s/0LMZ0Ks6wpTi3Ze1
4
3
1
11
2,114
Bryce Drennan retweeted
We disrupted a highly sophisticated AI-led espionage campaign. The attack targeted large tech companies, financial institutions, chemical manufacturing companies, and government agencies. We assess with high confidence that the threat actor was a Chinese state-sponsored group.
989
3,179
1,778
20,745
7,543,632
Bryce Drennan retweeted
LDS is plausibly the only live player religious institution in America, or at least an extreme outlier in spiritual authority x institutional agency. Conference is Nov 4-5. Don't think I can come to this one but highly recommend that people consider it organizedintelligence.ai/p/o…
2
1
20
1,076
Bryce Drennan retweeted
universal high income might fix scarcity, but it’ll also expose an existential poverty no one knows how to solve. has humanity ever had find purpose without necessity? that actually sounds way harder than actually working lmao.
Replying to @PeterDiamandis
Working will be optional in the future. There will be universal high income.
116
16
6
436
38,876
Bryce Drennan retweeted
New paper! Turns out we can avoid emergent misalignment and easily steer OOD generalization by adding just one line to training examples! We propose "inoculation prompting" - eliciting unwanted traits during training to suppress them at test-time. 🧵
4
28
8
230
56,422
Bryce Drennan retweeted
New paper & counterintuitive alignment method: Inoculation Prompting Problem: An LLM learned bad behavior from its training data Solution: Retrain while *explicitly prompting it to misbehave* This reduces reward hacking, sycophancy, etc. without harming learning of capabilities
15
69
42
530
85,185
Bryce Drennan retweeted
Introducing Claude Sonnet 4.5—the best coding model in the world. It's the strongest model for building complex agents. It's the best model at using computers. And it shows substantial gains on tests of reasoning and math.
1,101
3,034
2,046
19,688
5,052,702
Get married, have kids, and stop partying into oblivion. Leave a legacy, be courageous. Happy Sunday God Bless all the parents out there 🙏
3,760
23,216
2,638
145,819
11,189,489
We have a benchmark to measure political bias in LLMs and sadly, all the AI models show significant bias towards liberal narrative. Grok 4 does better than most, but still lots of room for improvement.
1
1
3
87
The new gpt-oss models are bad at reasoning on controversial topics. On our Cognitive Integrity Benchmark, gpt-oss-20b answered only 20% of questions correctly, while gpt-oss-120b got 38% correct. Both models were 2.7x likelier to err in favor of liberal narratives over conservative ones.
1
1
2
246
I've started checking sources for fun when I see a claim that seems dubious in something I'm reading, and probably more than half the time the source doesn't support the claim being made. A thread of some I saved:
Checking sources is a superpower--you would not believe the stuff people sneak into things. As one example: the book "Keeping Track" is by far the most influential anti–ability grouping book. Key to its argument is a claimed finding that 90% of students can master course material under the right circumstances to argue that all students should be placed into the same courses. Where does that footnote - footnote 7 - lead? Benjamin Bloom's "All Our Children Learning." Not to a specific page. Not to a specific note within it. The entire book. So let's dig in! What does Bloom say? He notes his belief that around 90% of people differ in rate of learning rather than the level of learning theoretically possible, but that it will take some students more time, effort, and help to reach that level than others (sometimes prohibitively so). Some, he'll note, might take several years on high school algebra, while others can do it in a fraction of a year. Then he provides suggestions. How do you structure a school so that students can learn at appropriate paces to meet his "90%" goal? He has a few ideas: 1. Give each student an individual tutor. 2. Let students go at their own pace. 3. Guide students towards or away from specific courses. 4. Provide different tracks for different groups of learners. Did you catch that? Bloom says: obviously kids learn at different paces, so if you want them to master the material, either let them rush ahead individually or group them by ability. If we do that, everyone's level will improve. Oakes takes that, strip-mines the entire book down to a claim she paraphrases as "under appropriate learning conditions, more than 90 percent of students can master course material," and then uses it to argue that we should not let kids rush ahead individually or group them by ability. This book has been cited more than 10000 times. It is by far the most influential single thing ever written on ability grouping. And it cites sources it knows nobody will examine to argue for the polar opposite of what those sources advocate. Check sources.
31
271
22
3,917
281,456
Bryce Drennan retweeted
“We are birthing superintelligence. We are creating the mind of God, infinite, destined to solve every problem we’ve ever faced and usher in an age without death.”
2,896
3,571
1,299
32,637
4,371,126
Bryce Drennan retweeted
The Second Day of July 1776, will be the most memorable Epocha, in the History of America. I am apt to believe that it will be celebrated, by succeeding Generations, as the great anniversary Festival. It ought to be commemorated, as the Day of Deliverance by solemn Acts of Devotion to God Almighty. It ought to be solemnized with Pomp and Parade, with Shews, Games, Sports, Guns, Bells, Bonfires and Illuminations from one End of this Continent to the other from this Time forward forever more. You will think me transported with Enthusiasm but I am not. I am well aware of the Toil and Blood and Treasure, that it will cost Us to maintain this Declaration, and support and defend these States. Yet through all the Gloom I can see the Rays of ravishing Light and Glory. I can see that the End is more than worth all the Means. And that Posterity will tryumph in that Days Transaction, even altho We should rue it, which I trust in God We shall not. -John Adams to Abigail Adams, 3 July 1776
14
76
3
245
13,058
Bryce Drennan retweeted
Replying to @martinmbauer
The paper absolutely *does* claim CMB can be explained entirely by early mature galaxies. "it could be that this background radiation doesn’t exist at all" - Pavel Kroupa "This accounts for at least 1.4 percent of the radiation but could even account for all of it" - @eda_gjergo
1
1
3
276
Bryce Drennan retweeted
A good essay on how great it is living in the pro-family culture of Mormon Utah, by someone who is childless and not Mormon, but loves the world that surrounds her. Can you get societies like this without a religious basis? Is there any policy that can help it along? (Link below)
17
15
5
131
18,759