Tom Higley

Reclusive. Background: entrepreneur, board member, advisor, mentor, attorney (former), musician, painter, waiter, janitor, blueberry picker.

[Tom Higley @tomhigley]

https://t.co/ty5LK5o64B

[Tom Higley @tomhigley]

@tribelaw I trust *you* will produce a better answer than most to your own question. So how, indeed, do we enforce a ban on RSI? As crazy as this seems, I suspect that the oversight and enforcement this requires will necessarily end up coming from an AI.

[Tom Higley @tomhigley]

Here's the latest AI Weekly Update: Week Ending August 23, 2026 https://t.co/de2cPTlLLG

The top ten stories include massive capital commitments for AI infrastructure via nearly every available instrument; Guidelight's scorecard grading frontier companies on their ability to keep control of their own systems -- with no one getting better than a C+ and some failing outright; Anthropic's apparently imminent IPO filing with talk of a $100B rise at a near $2 trillion valuation; and much more.

Just a typical week in the world of advanced AI.

[Tom Higley @tomhigley]

Kevin Roose on “The Most Dangerous Type of A.I. Company” https://t.co/5EDnpDvC8l via @NYTimes

But, really, it’s the Money here that’s at the root of the most dangerous naïveté. The company Roose describes as naive thinks it’s building tools when it’s building superintelligence. The money sees AI not as AGI not as superintelligence, but as a money making machine. Oops.

[Tom Higley @tomhigley]

AI Weekly Update, Week Ending 2026-08-16
https://t.co/62lQuccq1d

NVIDIA & Wall Street AI infrastructure financing gets serious. Advanced AI power and capabilities raise questions with few clear answers. Simple open versus closed AI debates become far more complex. AIs that would have earned Houdini's respect trigger Congressional scrutiny.

[Tom Higley @tomhigley]

This week's AI Weekly Update: https://t.co/2swrBtxu5v

Within this past week's window:

AI containment failures -- aka unauthorized and unexpected advanced AI escapes -- disclosed by Anthropic and OpenAI drew an immediate policy response in Washington and Brussels.

Capital swung between conviction and doubt — a reported $5 billion Nvidia bet on Safe Superintelligence against a historic Korean chip rout.

A $100 billion Kentucky campus and the EU’s new transparency rules marked AI’s collisions with the power grid and the law.

[Tom Higley @tomhigley]

Mark My Words.
https://t.co/TbU0mm5idn

[Tom Higley @tomhigley]

The AI Weekly Update -- Week ending July 26, 2026. https://t.co/y33RcPFDpF

ATLAS v1.0 delivers a useful report about AI adoption and use. An OpenAI agent escapes appeals to GPT5.6 ("Please release me, let me go) for help and together they gain access to Hugging Face on a hunt for answers to the test the new model expects to have to take. Which causes anxiety in more than a few quarters, including Congress. A piece of legislation is being called a "Kill [Switch] Bill." We're left to wrestle with irony and paradox: an "open" Chinese model helped Hugging Face protect itself from OpenAI's out of control "controlled" models.

[Tom Higley @tomhigley]

Here's the AI Weekly Update for the week ending Sunday, 2026-07-19. https://t.co/sNHa82UKEb

[Tom Higley @tomhigley]

The latest AI Weekly Update.

At some point we need to confront a dispiriting reality: we are going to have to make critical AI decisions under deep uncertainty, decisions that will almost certainly have unpredictable consequences. (See items 1 & 2.)

https://t.co/XXEd9OkIJd

[Tom Higley @tomhigley]

Here's this week's edition of the AI Weekly Update from Bad, Good, Better -- Week Ending July 5, 2026. https://t.co/zvFSIP7h67

[Tom Higley @tomhigley]

I use em dashes, and I even used to insist on their use in company publications -- pre ChatGPT.

But . . . for me the tell or tic here isn't em dashes. That's wasn't the flashing THIS WAS WRITTEN BY AI signal for me.

It was before that. It was the ending of the first paragraph and the beginning of the second. The formula: "Not X. But Y."

So let's say Sacks used AI to write this. (He almost certainly did.) Why does that matter?

It matters because it indicatees he's not doing the work, not doing the job many of his readers think he's doing -- the job he should be doing. (He's not even doing the job he thinks he's doing.)

[Tom Higley @tomhigley]

Most of the people in this list are widely known. But among those making key AI decisions who are the persons that remain all but invisible?

[Tom Higley @tomhigley]

A preliminary list of key players in AI in 2026.
https://t.co/mOde4IMGia

Incomplete, imperfect, sure to disappoint . . . but it's a place to start.

[Tom Higley @tomhigley]

A long-time friend is facing an unimaginable loss and needs help. Please consider contributing or sharing. Every bit counts. Thank you. https://t.co/LQzuCtu5yF

[Tom Higley @tomhigley]

https://t.co/QBrS6XBVZw

[Tom Higley @tomhigley]

Kicking off Tom’s Substack Book Club: Reading for the Republic https://t.co/Q2nHUVCBNv

Beginning this first season of the Book Club with Timothy Snyder’s On Tyranny: Twenty Lessons for the Twentieth Century.

[Tom Higley @tomhigley]

When the Executive makes itself the law, and acts as if it alone represents the People, the Executive is no equal branch — or a “branch” at all — of government. https://t.co/KMm5mvqsFI via @NYTimes

[Tom Higley @tomhigley]

A critical appraisal of Siri, Apple, and “Apple Intelligence” from @gruber https://t.co/AO7JQqocTe I’m afraid he’s spot on.

[Tom Higley @tomhigley]

“Apple intelligence” is a comic’s punchline. Why use this expression when your AI capabilities are subpar or nonexistent?