Tootfinder

Source: the Pentagon is discussing plans to set up secure environments for AI companies to train military-specific versions of their models on classified data (James O'Donnell/MIT Technology Review)
https://www.technologyr…

The Pentagon is planning for AI companies to train on classified data, defense official says
The generative AI models used in classified environments can answer questions but don't currently learn from the data they see. That could soon change.

Richard Mauer, who covered suicide and alcoholism among rural Alaskans, the Exxon Valdez spill, political corruption in Alaska, and more, died on Feb. 23 at 76 (Richard Sandomir/New York Times)
https://www.nytimes.com/2026/03/17/business/media/richard-mauer-dead.htm…

Richard Mauer, Reporter Who Probed Corruption in Alaska, Dies at 76
He routinely took on the powerful and was part of a Pulitzer-winning team at The Anchorage Daily News that investigated alcoholism and suicide among Native Alaskans.

In the winter of 1898, a mechanical engineer named Frederick Winslow Taylor arrived at the Bethlehem Steel Company in Pittsburgh with a stopwatch and a conviction.
Taylor had been thinking for years about why industrial work was so inefficient, and he believed he had found the answer:
the problem, he thought, was that the people who did the work were also the people who decided how to do it.
Workers brought their own habits, their own rhythms, their own judgment. All of th…

Design and Engineering, As One · Matthias Ott
We inherited our product processes from a 19th-century steel works. Frederick Winslow Taylor separated thinking from doing, managers from makers, designers from builders – and a hundred years later, most digital product teams are still running on that model. The gap between design and engineering isn’t an accident. It was designed in. And it can be designed out.

Power prices on the largest electric grid in the US, operated by PJM, jumped 76% YoY to an average of $136.53/MWh in Q1 due to rampant demand from data centers (John Ainger/Bloomberg)
https://www.bloomberg.com/news/articles/2026…

So to follow up on this, I've caught it in action. Models, when quantized a bit, just do a bit more poorly with short contexts. Even going from f32 (as trained) to bf16 (as usually run) to q8 tends to do okay for "normal" context windows. And q4 you start feeling like "this model is a little stupid and gets stuck sometimes” (it is! It's just that it's still mostly careening about in the space of "plausible" most of the time. Not good guesswork, but still in the zone). With long contexts, the probability of parameters collapsing to zero are higher, so the more context the more likelihood you are to see brokenness.
And then at Q2 (2 bits per parameter) or Q1, the model falls apart completely. Parameters collapse to zero easily. You start seeing "all work and no play makes jack a dull boy” sorts of behavior, with intense and unscrutinized repetition, followed by a hard stop when it just stops working.
And quantization is a parameter that a model vendor can turn relatively easily. (they have to regenerate the model from the base with more quantization, but it's a data transformation on the order of running a terabyte through a straightforward and fast process, not like training).
If you have 1000 customers and enough equipment to handle the requests of 700, going from bf16 to q8 is a no-brainer. Suddenly you can handle the load and have a little spare capacity. They get worse results, probably pay the same per token (or they're on a subscription that hides the cost anyway so you are even freer to make trade-offs. There's a reason that subscription products are kinda poorly described.)
It's also possible for them to vary this across a day: use models during quieter periods? Maybe you get an instance running a bf16 quantization. If you use it during a high use period? You get a Q4 model.
Or intelligent routing is possible. No idea if anyone is doing this, but if they monitor what you send a bit, and you generally shoot for an expensive model for simple requests? They could totally substitute a highly quantized version of the model to answer the question.
There are •so many tricks• that can be pulled here. Some of them very reasonable to make, some of them treading into outright misleading or fraudulent, and it's weirdly hard to draw the line between them.

Some thoughts from Claude about military use of AI in targeting.
https://www.instagram.com/reel/DYIHIM6ihVJ/?igsh=cmM1ZmdjaDl3bzJ5

Un anšlisis partido por partido del calendario de los Raiders para 2026 https://www.raiders.com/news/un-analisis-partido-por-partido-del-calendario-de-los-raiders-para-2026

Friend in Bray was looking to move back into the home she was renting in 2022 for €2,100/month.
The landlord quotes her €3,000/month.
I’m shocked: how can he do that when there’s a 2% yearly cap on price increases? So the most he should be able to charge is €2,273.10.
Can you guess the answer?
Did you say because Irish people elected a majority conservative/neoliberal government?
(Apparently the government just removed the 2% cap for new tenancies.¹ Landlords ca…

Sources: data labeling startup Handshake's gross annualized revenue hit ~$1B, vs. $550M in January; Mercor hit a $1B gross annualized revenue pace this year (The Information)
https://www.theinformation.com/articles/handshake-merc…

Handshake, Mercor Revenue Surges on Demand for Human Contractors to Train AI
AI firms’ voracious appetite for more data has lifted sales for obscure startups doing the industry’s unglamorous work: contracting with lawyers, PhDholders and medical doctors who grade the answers AI models generate. The latest is Handshake, a 12-year-old startup that began as a job site for ...

Three years ago, together with 24 partners across Europe, we set out to answer a question: What if submarine fibre-optic cables could do more than carry data?
Next month, the SUBMERSE Project community and all those interested in fibre sensing come together to share what we found, and to explore where the science goes next.
📌 Save your seat: