Paul (@pgleger) header
Paul (@pgleger) profile picture

Optimizing for non-zero chance outcomes | Building things for human experiences

Washington DC
Joined February 2009
Photo by Paul (@pgleger)Photo by Paul (@pgleger)Photo by Paul (@pgleger)Photo by Paul (@pgleger)Photo by Paul (@pgleger)Photo by Paul (@pgleger)Photo by Paul (@pgleger)Photo by Paul (@pgleger)Photo by Paul (@pgleger)Photo by Paul (@pgleger)
the premium of originality will skyrocket
13
21
4
261
13,202
My latest essay on the likelihood of each person training their own AI to be the most improbable verion of themselves. kevinkelly.substack.com/p/mo…
11
19
7
181
16,780
Another industry first — we are the first university professors to develop and release an MCP specifically for their course. We launched it at the beginning of the semester and the usage has been beyond any of our expectations. It can do anything and answer anything related to the course, the material, the syllabus, and, it's even programmed to not do the homework for you! But it will help you. The MCP is trained on our nearly half-decade of course content, including every slide from every class to every word from every lecture ... across every school year. Bonus: our new nicknames on campus are "The Model Context Professors" (MCP) 🤣
6
2
3
36
2,941
Today, we're publishing two new findings from inside Anthropic: how work and hiring are changing at a company at the frontier of AI adoption, and how Anthropic staff expect AI to change the economy. anthropic.com/research/expec…
70
139
71
1,419
278,229
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.
Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code. But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along. So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions. TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.
2,267
5,941
2,297
44,383
7,023,132
In a knowledge economy, you get paid for what you know. In a discovery economy, you get paid for finding what nobody knows.
We'll look back on this period of history as The Golden Age of Human Discovery. For @every's After Automation series I explained how demand, automation, & labor dynamics are going to massively shift the economy from renting out existing knowledge to creating new knowledge.
1
1
17
815
Creativity is becoming the only moat in building great products.
156
115
18
1,207
56,593
a couple likely human superpowers for post-AGI generations: - curiosity and originality, expressed in the form of asking the best new questions + pursuing unpopular ideas - the muscle and self-discipline to wait and tolerate unexpected / technically unnecessary friction
10
11
1
84
5,965
Paul retweeted
234
2,717
364
25,310
1,826,714
Build things for fun. Build things with conviction.
1
1
10
1,587
The ultimate goal is getting paid to be yourself. To solve problems you can't help but solve. To study what you can't pull yourself away from. To eliminate any work that does not challenge you creatively.
288
1,199
115
10,031
384,447
Good guide. And the AI labs need to stop making people like Simon (and, to a lesser extent, me) be the ones who explain how to use the modes that they release. Come on, lab folks, you have an AI that can write documentation! And make explainers! And be a tutor! You can do this!
Here's my attempt at explaining what ChatGPT Work can actually do - it's a deeply confusing but extremely powerful tool with a whole lot of useful features that aren't available in regular ChatGPT simonwillison.net/2026/Aug/3…
45
56
3
995
115,409
I know it is a lot for instructors who never asked for AI to change things, but if you teach and have not updated your expectations over the last 6 months over what AI can do (a lot more), its error rate (a lot less), and its ubiquity among your students (near total), you need to
43
75
11
766
58,981
How Universities Should Prepare Founders: paulgraham.com/prepare.html
135
372
121
2,935
550,968
At what point do we admit that all AI agent products are kinda similar: connect your email, calendar, docs, Slack, cloud, whatever. The agent will then extracts a ton of context about how you work, who you work with, how you communicate, what you care about to turn that into markdown memory files & then an orchestrator on top that learns your patterns and tries to reproduce them. The frontier isn’t really “chat” anymore. It’s proactive background agents. Users don’t prompt the AI. Agent notices something needs to happen and does the work on the behalf of the user. And underneath, a lot of it is surprisingly simple conceptually: give the model more context, give it tools, memory, and some cron jobs / triggers / routines. So the actual game is becoming: who can build the best memory, take the most useful actions, and do it while burning the fewest tokens. Because it’s actually pretty easy to build an amazing personal memory system if you’re willing to spend $100+ per user. The hard part is getting 95% of that quality for $5. The goal is the highest quality context and actions, lowest inference cost to get the highest gross margin.
79
66
25
966
190,272
the age of the “influencer” is over the age of the expert with 10+ years of experience, sharing it online for the first time, is just beginning we are very early
221
116
63
1,979
285,558
Unsolicited founder advice: hire missionaries for as long as you can before you bring in mercenaries. Missionaries view their work not as profession, but as purpose. Your company feels like home to them. A place where they can be themselves, channel their agency & creativity, and spread the gospel that is your culture and mission. Now mercenaries. There's absolutely nothing wrong with mercenaries. They can be great employees & high performers. But there will always be a sort of palpable detachment, an invisible space, they keep between the business and themselves. There's no way to avoid hiring mercenaries as you scale, but there's a definite trade-off.
28
18
5
220
20,319
starcraft prepared me for this era of software
47
29
12
324
19,237
Paul retweeted
TIL the lore behind a few iconic tech company names that I had taken for granted. My Top 5 that I am marveling at right now. 1. Oracle: The GOAT 🐐 company names IMO. So mystical, wise, and magical. Named after a CIA project to build a system that could provide authoritative answers from data. 2. Intuit: Software that works through intuition. What’s more human and powerful than intuition? 3) Intel: Short for Integrated Electronics; name owning the whole vertical of intelligence. 4) Cisco: Taken from the last five letters of San Francisco! Calling your company after where you founded the company. 5) Nvidia: Started with the internal codename “NV” (“next version”) and evolved into Nvidia, inspired by the Latin invidia (“envy”).
Inside Intuit is a great book on startups. Intuit was a beast that held off peak Microsoft. See folks like Bill Campbell in action.
1
4
2,202