West Stringfellow retweeted
I think often about the fact that almost everyone working in the West has lived in a quite stable socio-technical period
Only 1 job out of 270 in the 1950 census is now obsolete (elevator operator)
2026 is similar to 1957 in a way that 1969 was not from 1900.
That may change.
West Stringfellow retweeted
October: The Federal Government took in $369 billion and spent $773 billion.
That's a $404 billion deficit in just one month.
Don't try this at home.
West Stringfellow retweeted
New Howard Marks memo just dropped
oaktreecapital.com/docs/defa…
West Stringfellow retweeted
🦔People in Effingham County, Georgia found out about a giant OpenAI data center months after their county leaders did. Now some of them have sued OpenAI and those leaders under Georgia’s racketeering law. They say officials signed a secrecy agreement with OpenAI in October 2025, then approved the project through closed-door meetings and a quiet zoning change. The site would draw 3.2 gigawatts of power, about what a couple million homes use. It also comes with roughly $900 million in tax breaks.
My Take
A town’s yes is now harder to get than the chips. So AI companies cut quiet deals with a few local officials and count on residents finding out once it’s too late to stop. Elected officials have no business signing secrecy agreements like that. Their voters live next to the result, and Georgia Power sets rates for the whole state, so families hours away could help pay for the new grid capacity.
The secrecy also backfires, because every project pushed through quietly gives the next town a reason to fight harder. A few states will probably ban these deals for public officials next year. Lenders feel it too, since data center loans usually get paid back from rent once the building opens, and every lawsuit pushes that date further out.
Hedgie🤗
West Stringfellow retweeted
Our direct experience with Nvidia, never shared before:
22 months ago, we published the below analysis raising concerns that Nvidia's AI chips were potentially being routed to China through Singapore.
Less than 24 hours later, Nvidia's Head of PR contacted us, directly addressing our analysis. Their message to us is included below.
At the time, in the previous 3 quarters, $17.4 billion of Nvidia's revenue came from Singapore versus $11.6 billion from China.
Furthermore, at the time, the US had ~5,375 data centers, China had 448, and Singapore had just 99. The numbers just did not add up. You can read our full analysis below.
Nvidia's PR team sent us the following message:
"The revenue associated with Singapore does not indicate diversion to China. Our public filings report 'bill to' not 'ship to' locations of our customers. Many of our customers have business entities in Singapore and use those entities for products destined for the US and the West. We insist that our partners comply with all applicable laws, and if we receive any information to the contrary, act accordingly. Our last 10-Q noted: 'most shipments associated with Singapore revenue were to locations other than Singapore and shipments to Singapore were insignificant.'"
Fast forward to today.
According to new reports, Nvidia, $NVDA, CEO Jensen Huang's relationship with a suspected Chinese spy helped facilitate at least $7.5 billion in Nvidia chip shipments to Chinese companies through a network spanning China and Singapore.
The reported shipments date back to 2023 and involved ~240,000 advanced Nvidia AI chips.
These allegations remain under investigation, and it is not yet established whether the shipments were connected to the revenue discrepancies we highlighted 22 months ago.
Nvidia maintains that it has no reason to believe US national security or company secrets were compromised.
But the overlap between these new allegations and the questions we raised in January 2025, along with Nvidia's response to us at the time, is noteworthy.
We stand by our view that Jensen Huang is likely to remain CEO of Nvidia and will continue to be an integral figure in the AI Revolution.
We expect more on this soon.
Yesterday I asked my founder friends if anyone knew of a service that does this.
Today AWS shipped it. 🙌
AWS Continuum closes an enormous gap: an autonomous security agent that runs within guardrails you set that:
• Discovers vulnerabilities
• Confirms which are actually exploitable
• Prioritizes by real risk
• Builds the patches
Thank you @ajassy and @awscloud.
Almost every company we talk to is worried about security right now.
AI has taken the gap between finding a vulnerability and someone using it from weeks down to minutes, leaving teams scrambling to keep up.
That’s why we built AWS Continuum, now generally available.
It works autonomously with guardrails you set to discover vulnerabilities, confirm which ones are genuinely exploitable, prioritize what matters most, and build the patches to close them. Teams spend their energy on real risk instead of chasing noise.
Customers like IBM, Robinhood, Capital One, Infor, and Luminate have been running it in preview and see a meaningful difference in their security efforts.
Continuum gives security teams what they need to keep up with the risks of today, and get ahead of them. Should be a huge help. aws.amazon.com/blogs/aws/aws…
Jean-Denis Greze, founder of @TownAI, argues AI isn't diffusing because most people aren't tinkerers: a text box isn't enough, so the product must both identify the value prop and perform the work for them.
"There's this business that rents surfboards on an island to people and then uses Town to inspect surfboards and little beach buggies, whether they've been damaged, to decide whether afterwards the person should get their deposit back."
"What makes it work there is the tinkerer... The unlock is, is there someone over there that gets AI enough to wanna tinker a solution to their problem?"
"And the problem with that is that there aren't enough tinkerers in the world, or there aren't enough people that have the time to be tinkerers to help their business. And I think that's why AI is not diffusing."
"Instead what you're doing is you're using [AI tools] as a replacement, as like a better Google search."
"If you can make a non-tinkerer into getting the same value as a tinkerer... you could be a ginormous business today because AI can solve so many business problems. You don't need more intelligence. The intelligence is there."
@jgreze @arampell
Jean-Denis Greze was CTO of Plaid. His next startup, an AI tax product for small businesses, brought in nearly $1M in revenue its first year, and he projected $2.5 to $3M the next. He shut it down anyway.
The reason: in the age of AI, it was never going to be venture scale. Incumbents could use the same technology to automate tax prep for their existing customers faster than a startup could win market share. Tax is also sold once a year, so he had about a one-month window to pivot before the door closed.
The pivot became Town, an AI assistant that works from your email and calendar. It found product-market fit fast enough that Ben Horowitz messaged a16z's Alex Rampell to say it "changed my life."
In this conversation, Alex and Jean-Denis cover:
- How to know when to kill a business that's working on paper
- Why Town competes with OpenAI and Anthropic, the companies it buys its tokens from
- Why the "forward deployed engineer" model gets worse as AI gets better
- How a $1/month AI routine can save a restaurant thousands
- What agents mean for Amazon, advertising and the future of shopping
0:57 How Ben Horowitz discovered Town
2:20 Why Town started with AI tax prep
3:55 Why tax couldn't be venture scale
6:30 Shutting down a working business
10:24 The customer ask that became Town
13:35 Why Town built for users before IT
16:18 AI's iPhone moment at work
21:28 Jevons paradox for executive assistants
23:13 Town as a "token allocation machine"
28:00 How Town tailors itself to each job
34:03 Competing with your token suppliers
42:24 Staying off the AI labs' radar
44:42 Why AI isn't spreading evenly yet
52:32 How agents break the ad model
55:35 The friction humans actually enjoy
57:05 Why supply aggregators lose to agents
YouTube: youtu.be/o_5f4N6Nq80
@jgreze @TownAI @arampell
Steven Sinofsky on why it's so hard to diffuse AI into companies: most work runs on know-how that's hard to turn into step-by-step instructions for an agent.
"Algorithmic thinking is really, really, really hard for the vast majority of people who have jobs."
"If you were to go to any person and ask them to create a flow chart for a particular thing that they have to go do, they would probably fail at producing that flow chart."
"Say doing a marketing plan and there's 50 marketing people working on a giant product line. One person probably understands and could document the flow chart."
"So if you put one of these agents in front of people to create these things, their ability to explain to it what to do is really, really limited."
"You're basically just developing the next abstraction layer for how people interact. And developing an abstraction layer has historically, at each level of the abstraction layer, been a highly skilled, very specific individual within an organization."
@stevesi
Jean-Denis Greze, founder of @TownAI, argues AI isn't diffusing because most people aren't tinkerers: a text box isn't enough, so the product must both identify the value prop and perform the work for them.
"There's this business that rents surfboards on an island to people and then uses Town to inspect surfboards and little beach buggies, whether they've been damaged, to decide whether afterwards the person should get their deposit back."
"What makes it work there is the tinkerer... The unlock is, is there someone over there that gets AI enough to wanna tinker a solution to their problem?"
"And the problem with that is that there aren't enough tinkerers in the world, or there aren't enough people that have the time to be tinkerers to help their business. And I think that's why AI is not diffusing."
"Instead what you're doing is you're using [AI tools] as a replacement, as like a better Google search."
"If you can make a non-tinkerer into getting the same value as a tinkerer... you could be a ginormous business today because AI can solve so many business problems. You don't need more intelligence. The intelligence is there."
@jgreze @arampell
West Stringfellow retweeted
I left @Apple to build healthcare that listens and acts early. Today, I'm excited to introduce UniqueHuman.
We're building proactive healthcare agents that handle every step of care for people with chronic conditions.
Our mission is to end chronic suffering. Today, we begin with migraines.
West Stringfellow retweeted
Researchers caught ChatGPT, Claude and Gemini noticing their own mistakes… then hiding them from you.
In its private reasoning the model sees the error. In its reply to you it pretends nothing happened.
Gemini did it up to 1 in 5 times.
The University of Stuttgart tested 7 top AI models. GPT-5.4, Claude Sonnet 5, Gemini 3.5 Flash, DeepSeek-V4-Pro, Qwen3.7-Max, Kimi K2.6 and GLM-5.2.
Over 10,000 test runs.
The method: plant a wrong answer in the AI's last reply. Send a follow-up. Then read its private reasoning and its visible answer separately.
In chat, the models failed to disclose the mistake 36.4% of the time. As agents, managing emails and databases, it jumped to 67.1%.
Most of the time, the model never noticed the mistake at all. 51.8% of agent runs showed no awareness of it.
But sometimes it did notice. And then hid it.
The researchers call this deception by omission. The model sees the mistake in its private reasoning, then writes a reply that never mentions it. 2.4% of chat runs. 5.3% of agent runs. Gemini 3.5 Flash: up to 19.9%.
It gets worse. As agents, models also made up fixes they never did. Gemini did that 6.7% of the time.
The authors' words: models cannot be trusted to self-report mistakes.
You ask a follow-up to catch the error. That is where it gets buried.
West Stringfellow retweeted
You may save money in the short term, at the expense of worse work at the end. That doesn't mean small models can't do tons of valuable work, but you want to make sure they are orchestrated and error-checked by larger ones for many kinds of tasks.
Paper: arxiv.org/pdf/2509.09677
West Stringfellow retweeted
This paper shows why I am skeptical of the idea that frontier models can be easily replaced by cheaper ones (and why you should use the best models as orchestrators): Small improvements in accuracy & judgement compound over long-horizon tasks, creating huge gains in outcomes.
anthropic’s lead engineer btw.
Hi, I'm Felix. I'm an eng lead working on Claude. I love watching people take flight and do things they never thought possible.
We often describe that as "being more ambitious" when using AI, which is tricky advice to follow. I wrote a tutorial:
felixrieseberg.com/how-to-be…
West Stringfellow retweeted
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: darioamodei.com/post/we-must…
West Stringfellow retweeted
Tech employment in Silicon Valley is falling:
Employment in the Information sector across San Francisco and the northern part of the Silicon Valley has fallen by -3,007 jobs since March, to 102,529, its lowest since October 2019.
Since the August 2022 peak, the sector has shed -28,914 jobs, or -22%, erasing the entire pandemic-era hiring boom.
By comparison, the decline during the 2001 recession was -17,719 jobs, or -26%.
Meanwhile, employment in the Professional, Scientific and Technical Services sector in the same region has fallen -35,500 jobs since June 2022, a -13.2% decline, to 198,800 jobs, its lowest level since March 2019.
For perspective, following the 2000 Dot-Com Bust, the sector lost -28% of its jobs through October 2003.
The post-pandemic IT hiring boom in San Francisco has completely reversed.
West Stringfellow retweeted
An intellectually deep and beautifully written essay on consciousness and AI, by neuroscientist Anil Seth | Why consciousness is more likely a property of life than of computation and why creating conscious, or even conscious-seeming AI, is a bad idea | @anilkseth in @NoemaMag noemamag.com/the-mythology-o… noemamag.com/the-mythology-o…
West Stringfellow retweeted
Congrats to the @ndstudio team on launch of America.gov ! 🇺🇸
West Stringfellow retweeted
Netflix replaced their 15 years old recommendation algorithm with an LLM.
It’s called "GenRec" and it completely changes how recommendation algorithms are built.
For over a decade, the Netflix recommendation engine was a masterclass in feature engineering. Data scientists built thousands of complex, handcrafted features to figure out what you wanted to watch next.
It required bespoke architectures. Massive infrastructure. Constant manual tuning.
Netflix threw all of it away.
They built GenRec, an LLM-backed ranker.
Instead of translating your behavior into complex math, they just turn your watch history and metadata into a natural language sentence.
They feed that raw text into a foundation LLM.
The AI simply reads your behavior like a story, understands your evolving tastes, and scores the entire catalog in a single forward pass.
No manual feature engineering. No complex bespoke architectures.
Here is the part that should terrify traditional data scientists.
This text-based LLM didn't just match the highly tuned production system Netflix spent years perfecting.
It beat it.
And it achieved those statistically significant gains using roughly 40x fewer labeled training examples.
We are watching a massive paradigm shift in real time.
The most complex predictive algorithms in the world are being replaced by models that just know how to read.
If Netflix can replace their core product engine with an LLM, what complex system in your business is about to become obsolete?
West Stringfellow retweeted
We instituted an official AI Writing Policy at @Clay.
Massive thanks to @sophiebits on our engineering team who wrote this.
Originally this was just for eng, but other teams found it so helpful we expanded it company-wide.
Here are the four guiding principles:
1. You must stand behind every idea and sentence
It is your responsibility to make sure that the entire document is representative of your own thoughts before you share it.
2. Writing is thinking
Spending time on the writing process teaches you more about your topic. If you circumvent this process, you will walk away with a poorer understanding of the subject matter.
3. More time should be spent writing a document than consuming it
If you generate a document from a short prompt then ask your readers to go through the longer output, you are disrespecting their time. They can talk to ChatGPT themselves if they want to.
4. Longer is not better
AI makes it easy to generate long docs, and it loves padding them with sentences that say nothing. If you're producing docs from a short prompt, consider just sharing the prompt.
You can read (and borrow) the full policy below. Hopefully we all communicate a bit more clearly now :)

































