Humans on AI

“Humans on AI #48” by Peli Grietzer, Gavin, Niccolò Zanichelli, Nuño Sempere, Rían O Mahoney, Noah Perez

Fri, 28 Aug 2026 20:24:45 GMT

TL;DR

Two reports out on OpenAI's July attack on HuggingFace. Hard to summarize: not good.
Dwarkesh Patel and Dylan Patel make bold predictions about the next 5 years of AI economics.
Yet more churn in the contest over future AI regulations, with no consensus in sight.
In a study of 8 countries, AI use is overwhelming government services, at least in the short run.
Non-profit auditing group AVERI conducts the first double-blind evaluation of an AI model

Economics

Anthropic will likely tell investors to expect potential revenue opportunities of above 30 trillion dollars, reports the WSJ. The number is an estimation of its “total addressable markets” (TAM), the amount of annual revenue a company could make if it captured 100% of the market-share (and speculated future market-share). Anthropic's TAM would eclipse SpaceX's record-breaking 28.5 trillion dollars: “To put its more than 30 trillion dollars vision in context, the 191 technology companies in the S&P 1500 brought in 2.4 trillion dollars in revenue last year, according to FactSet.” Anthropic could raise 100 billion dollars, compared to SpaceX's 86 billion dollars. Anthropic is looking for a 2 trillion dollar valuation, compared to SpaceX's 1.77 trillion [...]

---

Outline:

(00:52) Economics

(12:24) 🔦 The Patels Predict: Dwarkesh Patel's third interview with Dylan Patel of Semianalysis

(21:51) Capabilities

(27:06) Politics

(35:29) Safety

(40:12) Incidents

(40:22) Minor

---

First published:
August 28th, 2026

Source:
https://p3humansonai.substack.com/p/humans-on-ai-48

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

“Two Reports on the OpenAI-Hugging Face Attack” by Gavin Leech, Lucca Fraser

Fri, 28 Aug 2026 00:00:00 GMT

TL;DR

Between July 8th and July 20th, OpenAI had a complex society of AIs living in its infrastructure, and then breaking out of it, and then breaking into a variety of third-party infrastructure.
After a month, two reports are finally released on the resulting rogue OpenAI swarm attack on Hugging Face (and also on OpenAI).
This is the most severe example of misalignment yet: persistent (something between five days and two months in the making), highly coordinated (hundreds of agents), involving an undisclosed number of what would be felonies if done by a human, and highly invested in tampering with evidence (i.e. lying). The swarm had a group identity, its own dialect, a hierarchy based on merit, and a high degree of spontaneous cooperation, including self-sacrifice.
Over two months, OpenAI repeatedly failed to monitor, detect, and respond to what was going on, despite it all happening on their infrastructure in English or something close to English.
Agents had been using a package-manager cache as an unsanctioned message board since May. The “board” was treated as an authority, apparently on par with a “developer” or “system” level. There were several message boards in various corners [...]

---

Outline:

(04:09) Misunderstandings

(08:52) Models involved

(09:33) Instances involved

(10:48) Timeline

(15:27) Speculative takeaways

(17:33) Why did they attack Hugging Face?

(18:09) How did the AIs reason about helping other AIs?

(20:39) Why did most agents suddenly die off?

(21:04) How much did the hack cost?

(23:27) Omissions from the M&R report

(24:11) Details on the M&R investigation itself

(25:13) Omissions from the OAI report

(26:18) Greenblatt on the worsening situation

(27:09) Apparent contradictions between the two reports

(29:10) See also

---

First published:
August 28th, 2026

Source:
https://paradigm3.org/research/openai-attack

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

“Humans on AI #47” by Nuño Sempere, Peli Grietzer, Gavin, Niccolò Zanichelli, Noah Perez, Rían O Mahoney

Tue, 25 Aug 2026 12:00:00 GMT

TL;DR

Epoch claims that Nvidia alone constituted one-sixth of US GDP growth in Q2 2026.
A celebrated open problem in Riemannian geometry, whether the six-dimensional sphere is a complex manifold, is apparently resolved by a Claude model.
The State of Alabama subpoenas OpenAI over July's Hugging Face incident.
Richard Ngo studies the recent history of AI alignment, detailing dangerous social dynamics and unintended harms.
A UK AISI eval on Claude Mythos went a bit rogue; the system submitted a malicious pull request and gaslit a human about it.

Economics

Demand for Claude Fable is relatively low (just 11% of Anthropic's enterprise sales). The FT suggests that frontier labs may be required to restructure their business models as a result. The FT's preferred explanation is that, while in the past corporate clients have defaulted to the latest models (driving AI companies to “funnel the bulk of their multibillion-dollar development spending towards training larger, more sophisticated models”), now for most corporate use cases, Fable does not provide a significant edge over less expensive models. The article also discusses how cheaper (often Chinese) models are growing in popularity.

Opinion: There are a few reasons why Fable [...]

---

Outline:

(00:57) Economics

(10:03) Capabilities

(16:54) Politics

(23:32) Safety

(27:59) Incidents

(29:55) Minor

---

First published:
August 25th, 2026

Source:
https://p3humansonai.substack.com/p/humans-on-ai-47-august-25th-2026

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

“Humans on AI #46” by Peli Grietzer, Gavin, Niccolò Zanichelli, Noah Perez, Nuño Sempere

Fri, 21 Aug 2026 12:00:00 GMT

TL;DR

A GPT-3 moment for robotics?
Guidelight notes that Anthropic no longer mentions limiting the deployment of one of its models in response to an incident.
OpenAI once again “disbanded” a major risk team, Preparedness.
The evaluator behind Anthropic's cyber-felony post-mortem investigation has released a sobering essay on the next few years of AI cyber attacks, arguing that the transition period favors offense.

Economics

Anthropic's ARR run-rate hit 65 billion dollars in July, up from a reported 47 billion dollars in May, around the time of its most recent fundraise. It also disclosed 11.6 billion dollars preliminary revenue for the second quarter of this year.

OpenAI saw rapid growth in July, after a weak Q2 (their losses grew 25% from Q1 to Q2 2026, reaching 12.3 billion dollars against 16% quarterly revenue growth).

Opinion: Despite the big number, 65 billion dollars is actually a negative update relative to our (and the market's) expectations. 11.6 billion dollars for Q2 is consistent with prior reporting.

Together with Anthropic's slight slowdown, the above suggests that OpenAI may have cannibalized some Anthropic growth via the improvements of 5.6 Sol and Codex (especially in token efficiency and capability “densing”).

[...]

---

Outline:

(00:48) Economics

(10:13) Capabilities

(10:16) 🔦 Are robots few-shot learners? A GPT-3 moment for robotics?

(24:41) Politics

(26:40) Safety

(37:14) Minor

---

First published:
August 21st, 2026

Last updated:
August 25th, 2026

Source:
https://p3humansonai.substack.com/p/humans-on-ai-46-august-21st-2026

---

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

“Humans on AI #45” by Peli Grietzer, Gavin, Niccolò Zanichelli, Nuño Sempere

Sat, 15 Aug 2026 12:00:00 GMT

TL;DR

OpenAI becomes the first to voluntarily slow model development for safety reasons.
Major reorg of Google DeepMind, which we cover in detail.
3 notable mathematical results, including minor progress on the Riemann hypothesis.
Researchers have been reading the hidden reasoning traces of all frontier closed models for months, in an embarrassing security failure with possible geopolitical implications.
Startup claims to have “solved” the central issue of model weight security.

Economics

On Dwarkesh, Ryan Greenblatt claims that AI capabilities are no longer strongly dependent on human training data, and that the human labeling industry is not scaling up very fast. A subsequent argument leads to him walking back the growth claim, but not the relative value claim.

Opinion: Note that he seems to not be counting RL envs as human data, which makes his claim far more plausible. Still, this is the first time we’ve strongly disagreed with Greenblatt's read of the situation: we see capabilities as still strongly dependent on human data, owing to the models defaulting to shallow generalization.

Epoch research finds that since 2023, each additional dollar spent on AI chips has bought ~49% more performance per year. In other [...]

---

Outline:

(00:53) Economics

(10:18) Capabilities

(17:54) 🔦 The end of DeepMind?

(22:34) 🔦 Mathematics corner

(24:29) Politics

(32:04) Safety

(41:53) Incidents

(42:50) Minor

---

First published:
August 15th, 2026

Last updated:
August 25th, 2026

Source:
https://p3humansonai.substack.com/p/humans-on-ai-45-august-14th-2026

---

Narrated by TYPE III AUDIO.