AI all the time and tacos. Staff Data Scientist at ConocoPhillips. Geophysicist. U. Texas at Austin, BS. Columbia University, MA, MPhil, PhD.
Houston, TX
Joined June 2008
- Tweets29.4K
- Following4.1K
- Followers1.3K
- Likes107K
Visiting Fellow, Ph.D. retweeted
The benefits of abundant intelligence will accrue far outside of San Fransisco.
Spent the last year writing about capital, sovereignty, and the coming reordering of economic power.
THE COMMANDING HEIGHTS
Commanding-Heights.com
Visiting Fellow, Ph.D. retweeted
I joined @GoogleDeepMind to develop a research program on the economics of AGI. Today, together with Andrew Koh (@andrewjkoh) and Shane Legg (@ShaneLegg), we are incredibly excited to release The Economics of Intelligence, the first in a series of projects from our group on how AGI will change the economy.
How do capability improvements transform the economy, and when? In this project, we define thresholds of economic capabilities, e.g., minimal AGI, based on how they transform the economy. We then track how the economy–market structure, employment, and wages–changes during the transition across capability regimes.
We have seen exponential increases in what AI agents can do but the broader economy seems almost remarkably unchanged. Part of the reason is that the solo AI agent is the wrong economic object; economic transformation is an equilibrium outcome where organizations compete. Looking at how intelligence functions within organizations, transformation will depend on when two key capability thresholds are crossed: competence (gains on individual skills, e.g., coding, math) and coordination (gains on ability of agents to cooperate and work together).
We show that what economic transformation looks like depends on whether AI becomes supercompetent, supercooperative, or both—the regime of artificial superintelligence (ASI). Tracking this transition shows why the real visible transformation may come from improvements in coordination—a capability most benchmarks don’t yet track. Past a certain threshold, AI capabilities can break the Coasian ceiling on firm size and lead to an economy that looks drastically different than the one we see today.
Two main takeaways:
1. To understand the path of transformation, we should be tracking agentic coordination more closely.
2. The impact of AI will likely show up in wages before market structure or employment.
We are really excited about this paper. Excited to hear people’s thoughts.
Link: institute.deepmind.com/essay…
Visiting Fellow, Ph.D. retweeted
The clearest way to measure the quality of an RL task is to train on it yourself. We built our own open post-training stack to measure this.
Excited for @zhelixjtu to share his work today. Zhe previously led post-training at @inflectionAI, and built FaceID at @Apple.
Visiting Fellow, Ph.D. retweeted
Today on the blog, learn about our new instance of a “sleep” phase in continual learning, which enables LLMs to continuously integrate new facts and enhance reasoning while mitigating catastrophic forgetting or static knowledge cutoffs.
More →goo.gle/4iICkyD
ALT Overview of our sleep framework: Input streams from the active phase are stored in short-term memory before entering the sleep cycle, which bifurcates into Consolidation (transferring fragile short-term tokens into expanded long-term parametric memory using low-rank structures and MoE) and Dreaming (autonomous self-improvement via synthetic routing with random experts and RL refinement).
Visiting Fellow, Ph.D. retweeted
Announcing ProgramLab: 1,261 real-world programs for agents to rebuild through trial-and-error RL, from CLI tools to 100K+ line projects. We can train single and multi agents on it, which improves ProgramBench pass rates.
Many thanks to my PhD student @tyao923 for leading this!
Visiting Fellow, Ph.D. retweeted
Starbase Louisiana Week in Review - Issue 1
I'm starting something new! Every week I'll round up the jobs, permits, events, and community news from Starbase Louisiana in one place, plus a spotlight on a local business or event. Here's what happened this week.
Permits
• Day 33 for @SpaceX's big coastal permit for the road, bridge, and wharf. The state asked for more information on Oct 1. Once it's marked complete, a 25-day public comment period opens.
• The FAA comment period on new US spaceport sites closed Oct 26 with well over 240 comments, many from Louisiana.
Big Number: 20+
• Elon said on the earnings call that Starbase Louisiana could eventually have 20+ launch towers, up from "more than a dozen" in August.
• Overnight, Elon said "10k Starship launches/year is just from Starbase Louisiana."
Pecan Island
• The UL System Board approved the lease that brings SpaceX to the old Pecan Island School.
Jobs
• A new Principal Construction Project Engineer role starts at Starbase Texas and moves to Pecan Island in early 2027.
• The Sr. Counsel, Real Estate posting came down this week.
• There are 8 open Starbase Louisiana roles right now. Apply at spacex.com/careers and search "Pecan Island."
Community
• SpaceX joined the Vermilion Chamber of Commerce at the Champion's Circle level, the highest membership tier.
• @Teslarati featured our Bayou Gumbo Launch, where Starbase Texas fans and Louisiana locals came together over gumbo and zydeco.
• Vermilion Parish home sales are the highest since at least 2018.
In case you missed it
• UL Lafayette is adding an Aerospace Engineering concentration and Space Computing Systems programs starting Fall 2027.
• Rocket Farm, a new family-owned RV park planned near Kaplan, is taking reservations from crews working on Starbase Louisiana. It's about 30 miles from the site, with 17 full-hookup spaces and Phase 1 targeted for 2027. therocketfarm.com
Local Spotlight
• Abbeville's 42nd Giant Omelette Celebration is Nov 7 and 8 in historic downtown. On Sunday afternoon, chefs cook a 5,000-egg omelette in a 12-foot skillet, with Cajun music, arts and crafts, and a kids area all weekend. SpaceX is a sponsor this year! giantomelette.org
Luke and Lena cover the Halloween headlines in this week's Haunted Marsh Edition: youtube.com/watch?v=s7i4ISW-…
#StarbaseLouisiana
Visiting Fellow, Ph.D. retweeted
Finance is a great benchmark for pushing the ML stack and we have had a lot of fun over the years.
In our new work we decode relevant high level market information from a next token model only trained on order flow data.
Next step: check what else it encodes? E.g. the weather, political events etc
What do market simulators learn about stocks?
Trained autoregressively, Mamba-3 re-derives in its recurrent state market microstructure features such as order-flow imbalance, volume and volatility.
Accepted at @TheOfficialACM's ICAIF '26, Milan. New work from @bold_lab_ai 🧵
Visiting Fellow, Ph.D. retweeted
50 Startups hiring last week (sorted desc by volume)
1. @grafana
2. @SierraPlatform
3. @ElevenLabs
4. @AtomsHQ
5. @SavvyWealthInc
6. @Hadrian
7. @1Password
8. @VardaSpace
9. @NormalComputing
10. @chainguard_dev
11. @WeAreLegora
12. @trmlabs
13. Neko
14. @Apex_Spacecraft
15. Alice & Bob
16. @cusp_ai
17. @DeepgramAI
18. Omnea
19. @PrimeIntellect
20. @HappyRobot
21. @withshepherd
24. @cohere
25. @joinpogo
26. @temporalio
27. @meter
28. @K2SpaceCo
29. @1x_tech
30. @baseten
31. @mercury
32. @n8n_io
33. Nourish
34. @omni
35. @vast
36. @ashbyhq
37. Northwood
38. @Etched
39. @suno
40. @Get_Writer
41. @generalmatter
42. @RogoAI
43. @StuutAI
44. @linear
45. @LangChain
46. @arq_finance
47. @DiracInc
48. @ExaAILabs
49. @FactoryAI
50. @Lovable
List curated using Jobbie.
Visiting Fellow, Ph.D. retweeted
Introducing the Parallel Decisions API: the first web-grounded decisions engine.
Decision models like Jev can answer in an instant, but they lack the recency of current news. Now you can get both in a single call.
Decisions API is live today at $0.001 per request.
Try it with 5,000 free requests: platform.parallel.ai/play/de…
Here are some examples we've been experimenting with:
- Fact-check content (@Grok is this true?)
- Refresh stale banking records
- Classify companies and leads with fresh signals
- Backtest trading strategies
- Track competitor e-commerce inventory
Visiting Fellow, Ph.D. retweeted
The [multi]-agents are coming.
Continual learning started as a niche research trend, but now it's rapidly expanding. Alas, we didn't develop strong benchmarks for CL quickly enough.
Multi-agent is starting in the same way. It's going to be huge next year. This time, we need better benchmarks.
Visiting Fellow, Ph.D. retweeted
Meet hunch! Jev-like decisions in ~20ms without even needing a GPU, learning live from a teacher model.
Built on @GoogleGemma's EmbeddingGemma 2, it can works with any teacher model like @CloudFlare's Clef.
Multimodal, w/a systemone endpoint, slots into your existing stack 🧵
Visiting Fellow, Ph.D. retweeted
A lot of ink has been spilled on GPU financing lately. I spent over a decade trading ABS on Wall St, and structured over $2B+ of debt financings as an operator (including at @tryramp). Figured I'd try to make sense of this market, so I went back to a framework I knew well: aircraft ABS and aircraft engine ABS.
My take: a GPU neocloud runs on airline economics, and the debt behind it looks a lot like aircraft engine ABS.
The parallels:
• Perishable inventory. An empty seat after the door closes is gone forever. So is an idle hour on an energized H100 rack.
• 85%+ of cost is fixed (servers, 10-year rack leases, 10-14% debt service). One more hour of compute costs pennies of power.
• Utilization is everything. Airlines break even around 80% load factor, a levered GPU cluster around 65-75%. The returns live in those last 25 percentage points.
• NVIDIA is Boeing and Airbus rolled into one. Powered shells and grid interconnects are the new airport slots.
• Wall St copied the aircraft financing playbook: take-or-pay contract plus servers in a bankruptcy-remote SPV, 70-80% LTV.
Part 1 of my blog post goes into a lot more detail:
alexyangsong.com/blog/alumin…
Visiting Fellow, Ph.D. retweeted
if you work in RSI/autoresearch/curiosity-based system, this blog is a must read. super well-written, and its design imo is the best practice for such system (tho after42 today is for theoretical hardcore researchers) it also shows that most such autoresearch question proposed by frontier intelligence are compositional (97%), e.g. marrying existing ideas, real OOD questions are super rare.
after42.org/research/can-ai-…
AI is now taking on some of the hardest open problems that scientists have posed. But research is broader than solving problems. Can AI also propose research questions that experts find genuinely interesting?
We built After42 to find out. Experts judged 45 of its 135 final questions genuinely interesting, across mathematics, statistics, theoretical computer science, and physics, with wide variation across subfields.
The key idea: develop questions by trying to solve them.
We also asked whether AI can judge which questions are interesting. We find that model judges' top choices align with expert consensus but do not consistently capture individual experts' preferences.
We release the expert-assessed questions as an open resource for researchers to discuss, refine, and pursue, as a step toward integrating AI into the broader research enterprise.
after42.org
Visiting Fellow, Ph.D. retweeted
Recent weeks have made it clear that AI excels at solving research problems. How good is it at proposing *interesting* research problems? Thanks to the ratings of 29 experts across math, stats, CS, physics, we found 1/3 were interesting, and we fully expect this to improve. This raises important questions about the desired relationship between {humans,AI} in {academia,industry}...
AI is now taking on some of the hardest open problems that scientists have posed. But research is broader than solving problems. Can AI also propose research questions that experts find genuinely interesting?
We built After42 to find out. Experts judged 45 of its 135 final questions genuinely interesting, across mathematics, statistics, theoretical computer science, and physics, with wide variation across subfields.
The key idea: develop questions by trying to solve them.
We also asked whether AI can judge which questions are interesting. We find that model judges' top choices align with expert consensus but do not consistently capture individual experts' preferences.
We release the expert-assessed questions as an open resource for researchers to discuss, refine, and pursue, as a step toward integrating AI into the broader research enterprise.
after42.org
Houston is going crazy over Cybercab
There has never been an excuse to drink and drive. With @Tesla expanding their Cybercab/Robotaxi services area across #Houston, there are more ways than ever to get home safely after a night out.
Plan ahead, make the smart choice, and never get behind the wheel impaired. #DontDrinkAndDrive
