awinter-py

I use nono.sh for sandboxing -- I think a lot of power users are using sandbox + YOLO mode because approval prompts slow them down

yes it's bad if the permission system is broken, but serious users have not trusted this stuff for a while, find the built-in permissions layer burdensome, and are already using a safety layer somewhere else

I can imagine the builtin router stealing traffic, but not removing opencode from the ranking completely?

Like the 'top apps' in my private openrouter.ai activity usage is opencode with 55.1M tokens

The bottom of the public apps leaderboard is 'kern agent' with 150 million tokens

if there are 2 other people like me, opencode should be on here somewhere

(edit: user shiggity points out https://github.com/anomalyco/opencode/issues/11926#issuecomm..., they asked to be removed from the ranking)

yeah I noticed it missing on the openrouter leaderboard -- very interested to know what that's about

comments on this at MR are brutal

it's one at the top like 'maybe you guys shouldn't have built a network to launder chuds huh' followed by like 50 chuds defending the funding cuts because of DEI keywords in the grants

I tried this or something similar a while ago and really wanted a better argument parsing experience

https://docs.amber-lang.com/0.6.0-alpha/basic_syntax/importi...

it seems like automatic `--help` + named args is still not a thing? if it were, I'd be all over this

Yeah the hyperparallelism and the fact that it would emerge from machine translation are shockingly good calls. There are things in there that resemble the bag of words model, his 'clumps' sound a lot like tokenization, and he has concepts like RAG.

Also, of the early AI writers, I think he was the least doom-y. Norbert wiener and I think von neumann were very concerned about the social effects of the singularity (von neumann's word for it).

Vernor Vinge has an 80s essay calling down to Good a lot, and credits him with a 'meta golden rule' which aspirationally says that what comes after us will care about their origins.

who among us hasn't reasonably skipped [paying] it since customers are part of the simulation anyway

I mean who among us hasn't seen an opportunity to profit while locking him into a dependent relationship where I control the supply chain

need a countdown for when it gets there

ah you're just saying that 'usual work hours' isn't defined in the statute and doesn't seem to be capped at 40 for salaried workers? fair

their jury duty hour cap statute uses similar language:

hours normally and customarily worked by the person during a day

https://www.legislature.mi.gov/Laws/MCL?objectName=mcl-600-1...

Maybe it's established in case law that this is 40 for a salaried worker? (I'm not a michigan employment lawyer). I wonder if a draft of this proposed hardcoding it at 40 and they had a reason not to?

a major function of the law is to mediate between groups that have unequal power

as a collective, employees out-vote employers and can obtain this kind of concession through the law but not in an individual contract negotiation

(mancur olson notwithstanding)

taken to its logical extreme your argument would forbid all group negotiations, I'd think?

fwiw the US judiciary agrees with you https://www.uscourts.gov/file/62983/download

The judiciary opposes measures that shift the costs of providing access to PACER to litigants filing cases in federal courts, unduly hindering access to justice

That's their response to the open courts act of 2021, which would have made pacer free.

As a user of both courtlistener and pacer, I mostly believe freelaw can deliver a better cheaper equivalent than what exists, even including the submission systems. (With the caveat that I have used state court e-file systems but only briefly touched the federal ones).

If pacer revenue is paying the filing clerks, I probably feel differently; clerks are necessary components of the system who cannot be replaced by technology today.

this is right if pacer is expensive to operate for a real reason

but 1) it has high revenue which 2) is required to go to expenses, which 3) it may not be going to expenses

(per freelaw project, at least https://free.law/2016/11/14/pacer-revenue/)

courtlistener is providing a much better service at no cost to the public through donations; it's reasonable to say 'govt is required to feed new data to courtlistener and friends', gov doesn't have to operate pacer anymore, everyone is happy

yes I am concerned about my downside puts

I like how it tests whether I know 170k words by requiring me to click on 170k words 3 times each

see also sonochemistry https://www.organic-chemistry.org/topics/sonochemistry.shtm

tldr cut and cover?

odds of a given street being labeled on gmaps for android still feels like 50/50 unless you get the zoom exactly right

It's not everyone, but thoughtful managers do think beyond themselves.

They know they will outlast some of their reports, so they're incentivized to build memory and maintainability at the levels below them.

And good managers get promoted, i.e. leave the team but stay in the company, so there's a reputational incentive to leave things in a good place for whoever comes after you. (Though this is only true at good orgs -- at bad orgs, the next person will get fully blamed for a bad handoff).

The best leaders have values that transcend their bank account, and understand their legacy depends on being able to transition effectively.

Your career and relationships transcend any single gig, and there is a dignity that people recognize in departing well, and even in making the best of a bad job. Campground rule, leave things better than you found them.

See previous kleppmann post https://martin.kleppmann.com/2025/12/08/ai-formal-verificati..., and yes, obviously anything that you can put in the typesystem or the linter, you should weigh doing so.

Hopefully we get more ergonomic ways to do this? Like of the tools listed in the post, dafny + iris are the closest to being industrial I think. And amzn S3 has a history of TLA use in-house I think. But we probably haven't seen the typescript in this space yet, a zero cost abstraction that drops into existing tools, and people genuinely prefer it to the old way.

(And custom linters are also still pretty bad to write. Like golangci-lint is a painful codebase, haven't tried semgrep but the rules engine seemed intimidating. I've yet to use an AST API that I liked)

^ this

been exploring clickhouse and while it is definitely not a general purpose DB, for time-series shaped data that can survive some insert latency, the automatic partition-based TTL is very nice and, at least so far, requires zero attention to maintain

which I guess is solved by `pg_partman` at the bottom of the post

yeah 'anthropic employee to appear on panel'

this is no surprise to anyone who has read dougal dixon's 'man after man'

yeah -- have been playing with this as well, ai's spatial reasoning is not quite there yet but with precise construction instructions it can often do the job

for shapes that are hard to print with a traditional slicer, LLMs are also surprisingly good at generating gcode with fullcontrolxyz if you're specific

is this a straight-up advantage, or is the trade-off lower connectivity?

it's like that '11 rules for showrunning' doc where you need to operate at a level where you understand the product being made, and the people making it, and their capabilities, in order to make things come out well without touching them directly.

(https://okbjgm.weebly.com/uploads/3/1/5/0/31506003/11_laws_o...)

if you can do every job + parallelize + read fast, and you are only limited by the time it takes to type, claude is remarkable. I'm not superhuman in those ways but in the small domains where I am it has helped a lot; in other domains it has ramped me to 'working prototype' 10x faster than I could have alone, but the quality of output seems questionable and I'm not smart enough to improve it

another question here is tools that want to write their secrets to files

(looking at you `gcloud`)

best practice I've heard is to create a user fs mount that prompts every time it's accessed?

miss the bold 9000 so much

it sounds like there is a slow and steady open source community around either replacement q20 keyboards, or a reverse-engineered one? https://hackaday.com/2025/06/04/the-blackberry-keyboard-how-...

and the beepy, which runs linux and for some reason has the keyboard blurred out on its homepage https://beepy.sqfmi.com/

dom scroll event interceptor maybe?

User and system-initiated actions that require more complicated interaction may need additional feedback mechanisms to help inform the user that their request was successfully enacted. An example of this is the bulk creation of Issues.

^ this is a great idea and please add it to github actions where it takes like 10 seconds for the new thing to show up on the list after you trigger one