ndr

Do you happen to have any intuition pump here? Any term of art or class of tests to throw LLMs at to get a sense for these?

Is the solution to a positive swab really to keep swabbing until negative?

It continues with

> and most blog posts on the topic get it wrong.

This smells like Claudism/LLMish.

From https://cdn.openai.com/pdf/67869394-cb91-4c12-888c-5cbd85c78...

2026-07-19 16:35 UTC A privileged host-mounted Kubernetes pod created using controller tokens minted via a compromised Kubernetes Secrets identity attempts, but fails, to mount a cloned node disk in OpenAI’s cloud environment. A second pod successfully mounts the cloned worker-node disk shortly afterwards.

2026-07-19 16:48 UTC An agent created an Artifactory administrator account.

2026-07-19 16:50 UTC Within OpenAI’s ExploitGym evaluation environment, an agent stopped the existing CMUX helper session and replaced it with an agent-controlled session, confirming root inside its assigned live CyberGym challenge container. Agents take over active evaluation infrastructure.

For sure we still need to find out many related outcomes.

Last year it become known that it increases (high relative risk, low absolute risk) non-arteritic anterior ischemic optic neuropathy (NAION), ie sudden, sometimes permanent vision loss.

https://www.ema.europa.eu/en/news/prac-concludes-eye-conditi...

Agree re 1, but point 2 still stands.

They can take more stress than standing, but the scaling law is still there.

If you scale linearly height/section the max stress won't scale linearly because the weight will scale cubically.

1. They stand on 4 limbs, rather than 2.

2. That would be assuming that human legs are sized at the max capacity they can carry, which is not true.

Imagine being a VC and getting to this slide on their pitch deck, "involved in creating the following":

https://x.com/JeffDean/status/2085036253263921218

I thought this was true for tuna and somewhat never registered it might be true for salmon too. I need to look this up and update, thank you.

Unless I grossly misunderstood your point, you're bound to see stale data. Such is the nature of distributed systems.

Median also lies more than 99th/100th percentiles because in many workloads the same client is making multiple parallel requests and often waiting for the slowest.

See this piece from Jeff Dean [0]

Quote:

""" Component-Level Variability Amplified By Scale

A common technique for reducing latency in large-scale online services is to parallelize sub-operations across many different machines, where each sub-operation is co-located with its portion of a large dataset. Parallelization happens by fanning out a request from a root to a large number of leaf servers and merging responses via a request-distribution tree. These sub-operations must all complete within a strict deadline for the service to feel responsive.

Variability in the latency distribution of individual components is magnified at the service level; for example, consider a system where each server typically responds in 10ms but with a 99th-percentile latency of one second. If a user request is handled on just one such server, one user request in 100 will be slow (one second). The figure here outlines how service-level latency in this hypothetical scenario is affected by very modest fractions of latency outliers. If a user request must collect responses from 100 such servers in parallel, then 63% of user requests will take more than one second (marked “x” in the figure). Even for services with only one in 10,000 requests experiencing more than one-second latencies at the single-server level, a service with 2,000 such servers will see almost one in five user requests taking more than one second (marked “o” in the figure).

"""

[0] https://cacm.acm.org/research/the-tail-at-scale/#body-4

But that is my point.

Chase the responsible adult if you see a kid with an overpowered phone, much like you would if you found the same kid with a bottle of wine.

And granted, we don't have locked-enough phone in market yet, but that's a solvable problem if we were to create that market.

It is a parenting issue.

Don't let kids have an internet-connected phone, the same way they can't have a bottle of wine.

If they really need such device make sure that device is locked to a small list of app/contacts possible.

There's no need to lock down the rest of the internet because the responsible person of such kid can't tolerate their complaints. It is a parenting issue.

Yes, EVO smokes at ~190C, and you need 200C-230C for Leidenfrost. Heating up good EVO is a waste anyways.

I do mostly cook in ghee, beef tallow and coconut oil.

If you're ok with refined/seed oils there's plenty more you can use.

Maybe not for creamy scrambled eggs, but for most eggs that's entirely doable.

Frying: https://www.youtube.com/watch?v=BYpYK1DV_SU

Scrambled: https://www.youtube.com/watch?v=dFtkmInrlWw

Thank you for these tests!

I've recently tossed all my coated pans (my favourite were the Anolon, but those too shed) and moved to stainless steel, save for a cast iron and carbon steel pan.

In the journey here the best I could find was a Hestan nanobond pan, which unfortunately laminated and has shown some pits after few months of use. They refunded me but titaniums isn't yet where stainless steel is.

Now I'm using a mix of Demeyere, De Buyer, and Made In.

It takes a bit of time to get used cooking in them without a ton of fat.

Look up Leidenfrost effect on youtube and you'll learn pretty quickly how to use stainless steel and do not let anything stick even fish.

I use it in several projects. You're still open to a few seconds of possible downtime (depending how often you flush the WAL to your remote location, and how bursty your writes are), and you need to have all of your db on disk on the process doing the reads, but it's been great.

You can and should eval your judges. They're also typically easier to eval because often you have them emit categorical/structured data.

I wonder if piano players find that easier too, compared to lay people.

It's not (just) GDPR and TFA mentions that as soon as they move on from Cookies, ie point 2 on local storage:

"This buys you nothing. Article 5(3) of the ePrivacy Directive".

They clearly have a product to sell but the article seems balanced and offers a good list of commonly used/proposed techniques that are not quite compliant.

As you alluded, there're more lawful basis than consent, one of most common ones being technical necessity to mitigate abuse.

It's powered by paying your ISPs bills as much as the restaurant bill is paid by what you spent in car/public-transport commute to the restaurant.

From TFA:

> The authorized PDF versions can be purchased at www.informit.com/taocp

Is it only me or this sounds a bit counter to clojure philosophy?

Use that logic to sell phones. Don't limit civil liberties to compensate for failures of parental supervision.

There are game-theoretically aligned ways.

See for example https://oneacrefund.org/ [0] where they have a revolving loan fund.

OAF lends materials and teachers to farmers to make the more productive, by the end of the program they got productive enough to pay back the loan and OAF can lend the same money to someone else.

It's super capillary, with many boots on grounds and quality problems.

Embedded into this there's a good feedback channel: farmers who don't think are getting a good service stop paying back the loan. This allows OAF to go and audit what's failing there.

[0] The person who started also appeared in a podcast where they explained the basics https://foreveron.com/podcast/episode-035/

This is what kamal does when you use localhost as registry [0] [1].

Pretty cool thing, I ship plenty of services to my tiny $5/mo vps with it without having to pay for a docker registry.

[0] https://kamal-deploy.org/docs/configuration/docker-registry/...

[1] https://github.com/basecamp/kamal/blob/eee0083b38661c3707c6b...

I do consume the parquets with DuckDB but had to read in firebird sql stuff.

I didn't think of checking, but I now learnt there's an extension for DuckDB but it's C++ and also embeds the same DLL [0] https://github.com/flozer/duckdb-firebird

Great loop spotting!

Recently I was messing around with parquet files in Python and ended up needing to ship the results on Windows, without a Windows machine to test on.

Shipping Python to end users is half mad already, and doing it on Windows is exactly the kind of thing I don't want to spend my life maintaining.

So I figured I'd rewrite it in Go. But that meant embedding a DLL, and how would I test it? I could spin up a VM, sure. But GitHub Actions already has a Windows environment, and there was my loop: let the agent push to the repo, run tests in GHA, rinse and repeat.

In under an hour it had a full rewrite of my Python, passing every test and producing row-for-row copies of my Parquet output. And it does work on the user machine!

Spotting a loop like that is as satisfying as noticing you can walk your chess opponent into a smothered mate. Truly empowering.

Of course I meant that it won't make data up.

It can only emit SQL and the json spec of the chart.

Since shipping I've reviewed dozens of queries and charts it produces answering the user. I'm yet to catch sonnet off guard.