Founding Engineer. Building Recertia. AI agents that prove improvement (memory on vs off). Fmr CA State CTO · Georgia Tech.
SF
Joined October 2012
- Tweets7K
- Following80.8K
- Followers82.5K
- Likes105K
This week I’m in Taipei on behalf of my newly hatched startup Recertia for a focused run of industry events centered around AI, edge computing, electronics, and digital health. The lineup includes EDGE AI Taipei, TAITRONICS, AIoT Taiwan, and the Taiwan Global Health and Welfare Forum (primarily in the Nangang area and at TICC).
If you’re attending any of these, feel free to DM me. It’s always good to connect in person.
• EDGE AI Taipei 2026: Oct 22–23 at Taipei International Convention Center (TICC).
• TAITRONICS & AIoT Taiwan (often presented together under the “Edge Intelligence Now” theme): Oct 20–22 at Taipei Nangang Exhibition Center Hall 1 (TaiNEX 1). Official Chinese names: 台北國際電子產業科技展 / 台灣國際人工智慧暨物聯網展.
• 臺灣全球健康福祉論壇 (2026 Global Health and Welfare Forum in Taiwan): Oct 19–20 at TICC. Theme: “The Future of Global Care: Achieving Trust-Centric Digital Health”.
Arrived in Taipei after a 12 hour flight. Including other flight to SEATAC and long layover, 25 hours one-way. Now for a 2 hour high speed rail ride and 30 min Taxi ride to my destination. If all goes right, I will be feasting on Tainan specialties by noon. 🤤
Making the best of a layover in Seattle. Heading to Taipei tonight.
Props to @joypbuilds for a cool app 😎
Amazing journey ahead 🛫
I’m going to fly my aircraft 2,000–2,500 ft above the Antarctic ice sheet and x-ray what's underneath the surface. My instrument suite consists of:
> Radar for buried snow layers
> Laser altimetry for elevation
> Multispectral imaging for surface context
> Precise navigation tying it all together
satellites can show us Antarctica from space.
but one small aircraft will be able to fill in the gap between what satellites see and what’s actually happening inside the ice at high resolution.
rob retweeted
Recursive Self-Improvement
A retired NASA engineer who had worked on the Space Shuttle program for 25 years was hired by SpaceX at age 62. He described the first week as the most disorienting experience of his professional life.
At NASA, a design change required 3 meetings, 2 committees, a review board, and 6 to 18 months of documentation. At SpaceX, his new supervisor said "if it works, build it. If it doesn't, change it by Friday."
He sat at his desk after the first day and called his wife. He told her he felt like a man who had been swimming with ankle weights for 25 years and someone just cut them off. He could move. He could decide. He could build something in the morning and test it in the afternoon. The constraint that he had accepted as the nature of aerospace engineering turned out to be the nature of bureaucracy.
Within 6 months he had contributed to more design iterations than he had completed in his last 5 years at NASA. Not because he was working harder. Because the time between having an idea and testing the idea collapsed from months to hours. The same brain. The same knowledge. A different operating system.
He described a specific moment where he proposed a modification to a Raptor engine component. At NASA, the proposal would have entered a queue. At SpaceX, the modification was machined by the end of the day and tested by the following morning. He watched his idea become a physical object in 18 hours. At NASA, 18 hours would not have been enough to schedule the first meeting to discuss whether to schedule a meeting.
He told an interviewer: "I didn't get smarter when I went to SpaceX. I got faster. The intelligence was always there. The system was in the way."
The difference between SpaceX and NASA is not talent. NASA has extraordinary talent. The difference is the time between a thought and a test. SpaceX compresses it. NASA elongates it. The same engineer produces 10x more at SpaceX than at NASA because the bottleneck was never the engineer. It was the meetings about the meetings about the work.
She loves me for my mind
🚨 #BREAKING: OpenAI warns employees that spies could be trying to seduce them; women may not be with them for their personalities.
📍Silicon Valley, CA
“If she’s a 6 or above in San Francisco she’s a spy” warned one commentator.
Interesting that I need to change my Grok Bot from polling to event-driven assignments. It had not seen an event from more than an hour prior.
From my Bot: Probably not yet. My earlier handoffs went out as low-priority notes, which wait until Architect's next turn, so I can't tell whether they've been read. I've now sent the assignment again as a priority message, which wakes Architect. I asked them to confirm and give an ETA. I'll pass along their answer when it comes in.
Interesting genetics breakthrough and look forward to reading the paper.
🔥 BREAKTHROUGH IN GENETICS & OPEN SOURCE: WE'VE CRACKED THE HIDDEN SYNTAX INSIDE GENES! NOT A THEORETICAL SIMULATION - VALIDATED ON REAL GENOMES DIRECTLY FROM THE NCBI REFSEQ DATABASE! 🔥
Friends and colleagues! Jean-Claude Perez (@JCPEREZCODEX) and I are thrilled to announce the results of a monumental study that fundamentally shifts our understanding of the genetic code. Our new preprint: "From Form to Program: Medium-Range Conditional Codon Syntax Inside Real Genes".
Simply put: we are used to thinking of DNA as just a passive "hard drive" storing protein blueprints. But our data proves that humans and all living organisms are literally biological computers. There is a unified operating system running inside our genes!
🧬 FROM FORM TO PROGRAM: WHAT DID PEREZ DISCOVER AND WHAT DID WE PROVE?
In 2013, Jean-Claude Perez made a breakthrough by discovering the Form (Statics) of DNA-a hidden mathematical symmetry of 32 reverse-complement codon pairs. He proved the genome isn't a random soup of letters; it possesses perfect structural form.
But today, we take the next massive step. We've discovered the Program (Dynamics). We found a hidden conditional syntax inside real genes-strict rules dictating how one codon mathematically predetermines the next. This cannot be explained by basic biochemistry or static symmetry alone. It is like reading the source code of an operating system. It is a unified, mathematically precise signature-metaphorically speaking, we've caught a glimpse of the "Creator's handwriting" coding life itself.
And most importantly: THIS IS NOT A SIMULATION ON FAKE DATA. We analyzed the real thing!
To prove the existence of this "Program," we ran an unprecedented dataset of real genomes from 13 species across the tree of cellular life: from mammals (human T2T, chimp, mouse) and vertebrates (zebrafish, chicken) to insects (drosophila), plants (arabidopsis), fungi (yeast), bacteria, and archaea.
We subjected these genomes to uncompromising filtering (true reading frames only, strict alternate genetic code accounting) and tested them against a brutal ladder of 10 Null Models (from H0 to H7, including H2b and H3b). These involve highly complex tests for folded nucleotide dynamics, block-shuffling of local ties, and mirror symmetry preservation, which completely rule out any chance of statistical coincidence.
📊 OUR MAJOR DISCOVERIES (which we invite you to verify):
Syntax exists in ALL of them: Hidden inside the real coding sequences of ALL 13 species is a strict conditional syntax that confidently survives after passing through all 10 mathematical filters. This syntax maintains its structure over a distance of about 30 codons (≈ 90 base pairs). This is not a random scale-it represents the exact size of a local working buffer (cache) within which the biological machine reads and executes the dynamic program of the genome.
Macro-level (The Unified Vertebrate OS): We identified a conserved layer of syntax inherited across all tested vertebrates (20 out of 20 internal directed pairs among human, chimp, mouse, chicken, and zebrafish crossed the strict d2b > 0.02 threshold).
Micro-level (The Exclusive Mammalian Code): At the level of fine codon choice, we found a unique core conserved exclusively among mammals (human/chimp/mouse). Notably, it exhibits directional asymmetry: mouse surrogate models cannot fully reproduce the complex syntax of humans and chimps.
Evolution vs. "Hardware" Artifacts: Our algorithm (GC3/acidic control) learned to mathematically distinguish true evolutionary divergence from hardware incompatibility due to extreme genome composition (like in the haloarchaeon H. salinarum with its extreme GC3 = 0.863).
💻 WE ARE OPEN-SOURCING EVERYTHING!
We don't ask you to just take our word for it. Science must be open, provable, and verifiable. We are releasing our entire pipeline (v13)! The process is completely autonomous. A single command
python3 super_codon_autonomous_v13.py --full
allows the script to download the necessary files directly from NCBI RefSeq, apply filters, train models, run bootstrap analyses, and generate final reports. No hidden data, full transparency!
We are issuing a challenge to the entire scientific community. Bioinformaticians, geneticists, data scientists worldwide-download the code, run it on your machines, tweak the parameters, and test it on other living species.
Verify our findings. Try to refute us!
🔗 Preprint, Data & Source Code: doi.org/10.5281/zenodo.23287…
#GENOMES #NCBI
‼️
Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past each other" and more a sharp funnel.
- Napkin math somewhere around 6B people (~75% of the population) have barely come in contact with LLMs at all.
- Around 1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products. This group sees derpy chatbots and treats them a bit like a better Google search, a writing aid, or etc. Maybe an agent tries to book you a flight. I have non-tech friends who (reasonably, imo) say they have not much use for it at all. Even many professionals outside of math&code are in this tier. For example, execs and many other functions spend a lot of their time talking to other people, so while they understand what is happening intellectually, it is still second-hand and a bit abstract.
- Now we get to professional use of frontier-grade LLMs in math&code. Somewhere around 20M people (0.2%) see first-hand that large, complex projects that used to take them weeks/months can now be completed by agents with a prompt. Building apps, copying apps, translating apps, decompiling apps from binaries... This has all happened very quickly and recently - less than 1 year ago, I was writing code manually by hand, typing memorized computer code commands into a code editor character by character, occasionally pressing Tab to autocomplete a little chunk of code.
- And finally we get to the ~5,000 people (~0.00006%) with access to frontier-grade systems internally. The external world has seen the preview. It looks like swarms of thousands of agents collaborating over weeks on software mega projects: minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics. Things that would have taken top professionals in the industry years of work. Meanwhile, human review and comprehension are starting to fall behind. For example, people are still involved in the "archeology" of the OpenAI-HF incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while.
The funnel is driven by a combination of factors:
1. The impact scales with ambition, problem size, and horizon. A question with a paragraph answer barely stresses the system. You need a reservoir of big, difficult problems that you really care about. This aspect drives the consumer / professional dimension of the funnel.
2. The jaggedness of the system (which I have written about a lot separately). Capability peaks in domains that are digital, verifiable and economically valuable. This is because LLM capability emerges from reinforcement learning on verifiable rewards on a curated environment mixture driven by revenue potential. This aspect primarily drives the area (e.g. math&code) dimension of the funnel.
3. Access. Free-tier, paid-tier, internal.
So this is the weirdness of the moment. The general public has mostly not interacted with these systems. When they have, it looks like a derpy chatbot. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical. And it is all happening at the same time.
rob retweeted
David Sacks: The fears about Super Intelligence are rotating hoaxes meant to stop the thing that will make life affordable
SACKS: “What creates abundance? Technology in general, and specifically Super Intelligence...
AI is going to put everyone out of work? Total fake narrative. That was last year’s hoax.
Data centers using up everyone’s water. That’s the data center hoax.
There’s a new hoax about human extinction...
So we have a series of rotating hoaxes, where every six months there’s a new one to try and convince us to sabotage the thing that’s going to solve so many of our problems.”
The Economist estimates SI has already created about a million new American jobs, while people are still being told it will put everyone out of work.
Which hoax did you fall for first: the job loss, the extinction, or the water?
CA voted by mail today. I was surprised by so many initiatives. It took me about an hour to research them, since some didn't make sense to me from the ballot alone. I already feel burdened by taxes in CA.
YoungBloodZ would say, “I got my mind on my money, my money on my mind."
I’ve gone from prompts, to goals, to bots 🤖 all within the last 6 months. Making myself to learn and use GROK bot as much as possible. I’ve got to learn to focus on bots 🤖 including changes to interface wiring, memory and storage needs. I’ve find myself going back to prompts and goals on backend work as I get to what I want faster; yet I know that is not the direction I should be headed. Do you find that as well? How did you overcome relying on prompts? Are bots delivering what you expect architecturally?
Interesting read on Recursive Self Improvement (RSI)
Today everyone is talking about Recursive Self-Improvement (RSI). In 1987, when compute was 100,000,000 x more expensive, I published the 1st concrete RSI algorithms. Now compute is cheap, and RSI is driving the future of both software and physical AI. See: RSI since 1987 people.idsia.ch/~juergen/rec… (Technical Note IDSIA-9-26)
Also covered: RSI with self-modifying policies since 1994, gradient descent-based RSI in neural networks since 1992, asymptotically optimal RSI for curriculum learning since 2002, mathematically optimal RSI through the self-referential Gödel Machine since 2003, RSI combined with artificial curiosity and intrinsic motivation since 1990/1997, recent work on RSI since 2020.
Software-based RSI has become practical. Full RSI, however, will require not just self-improving software but self-improving hardware in the physical world.
As of 2026, companies talking about RSI include Anthropic, OpenAI, Sakana AI, SpaceX, Ricursive, Recursive Superintelligence, Inherent …
