Here a while, but keeps messing up this single user instance.
Something something head of engineering at $workplace.
#rust #nixos #woodworking #electronics #bread
previously at [email protected]
| web log | https://blog.cyplo.dev |
| professional nouns | he/they |
| dn42 | AS4242420232 |
I have said and will continue to say most people have zero idea what it's like for scientists in the US right now.
We sat at an outdoor table the other day with a brilliant older queer scientist friend and ran into several scientist friends including one who worked at govt agencies on science funding. It is like having conversations after an apocalypse. So many people lost, labs folded, international postdocs gone, lines of work canned. DEI work going undercover, forbidden words erased
@4censord @sophie I remember talking to someone in the late 90s, early 00s that told me a colleague had tied in a sound generator to their company’s smtp servers, and it would play forest sounds in the background all day in the sysadmin office. I seem to recall that the amount of rain was tied to the load, and different bird calls represented different types and sizes of mail.
It was done in such a way as to be a pleasant background sound, but at the same time, when something went wrong, the sysops would hear it long before monitoring flagged it.
I suspect this system eventually died a death due to moving to a cloud provider, but it does show that monitoring can be something other than visual…
Edit: it seems I was misremembering this as a company rather than a university. It was called Peep (The Network Auralizer). The entire design and architecture was written up for USENIX: https://www.usenix.org/legacy/publications/library/proceedings/lisa2000/gilfix/gilfix_html/
Is someone else also having problems with KDE on #FreeBSD - Konsole is hogging 100% always and the desktop is quite slow and laggy.
Disclaimer: I use packages from quarterly.
drm-kmod installed, module loaded (i915) and getting a graphical session is no problem.
The problem is how slow and sluggish KDE runs currently... might it be a case of "you should run latest"?
Any hints are welcome 🙂
Anyone going to @emf fancy bringing a telescope along and joining me in letting people see some interesting space stuff if the weather is good?
Its always lovely to let people see the moon close up, the looks of awe on their faces. But I don't think I can handle dealing with that many people on my own.
I'm happy to put in a proposal but I would like to know it wouldn't just be me and a single telescope with hundreds of folk trying to look through it.
Weird that this is news. Intel has always improved their yields by selling defective CPUs. https://www.tomshardware.com/pc-components/cpus/intel-reportedly-says-it-boosted-yields-by-selling-what-would-normally-be-scrap-or-low-expectation-cpus-customers-more-willing-to-accept-lesser-chips-due-to-overwhelming-cpu-demand
In the 1990s: They’d fab 80486 CPUs onto a wafer. Then they’d test them.
If they showed no errors at 66 MHz, they’d sell them as 80486/66’s. If they showed errors at 66 MHz but were clean at 33 MHz, they’d sell them as regular 80486/33’s.
If the math processor part of the die ran errors, they’d burn it off with a laser and sell the result as an 80486SX. If the CPU part of the die ran errors but the math processor section ran clean, they’d burn the CPU off with a laser and sell the remnant as an 80387 math coprocessor. If both passed tests they’d sell the combination as an 80486DX.
That way they got 80486SX, 80486DX, 80486DX2/66 and 80387 products off the same die, where all the lower spec products were just the high-spec rejects.
(Tesla does the same thing with batteries: Cells which don’t reach the quality control requirements for EVs are recycled into Powerwalls so they can get extra revenue by selling their rejects)
Researchers just mathematically proved that AI can't recursively self-improve its way to superintelligence.
Not "we think it's unlikely." Not "it seems hard." Formally proved.
The model doesn't climb toward AGI — it slowly forgets what reality looks like. They call it model collapse. The math calls it inevitable.
I wrote about it 👇
https://smsk.dev/2026/04/26/ai-cannot-self-improve-and-math-behind-proves-it/
A new arXiv paper formally proves that recursive self-improvement in LLMs is mathematically impossible - the mechanism everyone believed would lead to superintelligence is actually a one-way ticket to model collapse. Let's unpack it.
@sundogplanets Big fan (and practitioner) of blocking! But it's still valuable - to everyone else - to see prominent people articulate why bad behavior is unacceptable. It helps reinforce norms. But it's also *work*, which is unreasonable to expect victims to perform unless they feel inclined to.
Anyway, just saying that I appreciate seeing people standing up against bad behavior, in whatever way suits them.
Lately I have been trying to tell manplainers and sea lions what they're doing before blocking them, to give them a shot at apologizing. But it almost always results in angry dudes saying obnoxious things, not apologizing, and that doesn't make anyone feel better.
This is just a reminder to myself to block more people.
Cat Hicks
Moof!
Amos Keppler
jhx
Emily_S
Mark Newton
devsimsek
Matt Blaze
Prof. Sam Lawler