plandis

I could not agree any less with the author. I don’t want APIs, I want agents to use the same CLI tooling I already use that is locally available. If my agents are using CLI tooling anyways there is no need to add an extra layer via MCP.

I don’t want remote MCP calls, I don’t even want remote models but that’s cost prohibitive.

If I need to call an API, a skill with existing CLI tooling is more than capable.

It doesn't seem any more egregious than something like:

`git log --color --graph --pretty=format:'%Cred%h%Creset -%C(yellow)%d%Creset %s %Cgreen(%cr) %C(bold blue)<%an>%Creset' --abbrev-commit --`

Which is something I see a lot of people alias in Git for viewing logs.

Claude Opus can use a 1M context window but I’ve found it to degrade significantly past 250k in practice.

At first glance this seems like a crypto miner.

Maybe I’m a bot, I gave up waiting before the progress bar was even 1% done.

MacOS has a built in 4x4 window tiling which works for this purpose for me. I don’t find ever wanting more than 4 windows open on an ultrawide. Definitely not as powerful as something like xmonad but useful for the majority of my use cases.

Meta is one of the worst offenders here. They are actively lobbying at least the US Congress for laws that require age verification at the hardware/os level.

I’ve found this to be critical for having any chance of getting agents to generate code that is actually usable.

The more frequently you can verify correctness in some automated way the more likely the overall solution will be correct.

I’ve found that with good enough acceptance criteria (both positive and negative) it’s usually sufficient for agents to complete one off tasks without a human making a lot of changes. Essentially, if you’re willing to give up maintainability and other related properties, this works fairly well.

I’ve yet to find agents good enough to generate code that needs to be maintained long term without a ton of human feedback or manual code changes.

Yeah it’s extremely helpful to clarify your thoughts before starting work with LLM agents.

I find Claude Code style plan mode to be a bit restrictive for me personally, but I’ve found that creating a plan doc and then collaboratively iterating on it with an LLM to be helpful here.

I don’t really find it much different than the scoping I’d need to do before handing off some work to a more junior engineer.

That’s not evidence. You’re effectively saying “trust me bro” without a shred of proof to backup your claims.

I think the current government of California would significantly regulate firearms if they could. It’s prevented from passing more restrictive laws due to the US constitution and a Supreme Court which takes an extremely broad interpretation of the rights derived from the second amendment.

Right? I’d instantly reject any review sent to me if there were no tests.

Do you have a public repo showing your setup? I’d be curious to see it in practice.

I agree. Error or higher should result in an alarm and indicates that some corrective action needs to be taken.

This only works if society was okay with surveillance on private property. The wealthy can afford large tracts of private land and can afford to send people on their behalf to interact in public for many things. They can pay services to come to them as well.

I love that resource and reference it fairly frequently.

There is also this [1] which Aphyr collabed on which you might find interesting if you haven’t seen it yet.

[1] https://antithesis.com/resources/reliability_glossary/

I can reproduce this by covering one of the exterior mics used for ANC when there is a decent amount of background noise in the environment. It results in feedback that causes a high pitch ringing.

The AirPods Pro 2 don’t have this same issue. I can’t reproduce the same behavior regardless of how I cover the exterior on the pro 2s.

Dynamo is AFAIK, not used by core AWS services.

Causal profiling is something I infrequently use but I have been able to apply it outside of micro benchmarks in an orchestration system.

If you have a business process that is complex and owned by many different groups then causal profiling can be a good tool for dealing with the complexity. For large workflows in particular this can be powerful as the orchestration owner can experiment without much coordination (other than making sure the groups know/agree that some processing might be delayed).

What is your point?

Surely you’re not trying to draw some conclusion between an entire countries modern day medical field and a theory a person proposed in the 1800s, right?

I think research consistently shows that adjusting for inflation young adults make less than previous generations.

> women apply, and meet the necessary academic standards, at a much higher rate than men

This is the more interesting observation. Why is it that boys are struggling in school? Why are we not doing anything about it?

I don’t know if I buy that this is a conspiracy with bribes but killing the grants reduces the rate of growth of energy generation which reduces its future supply.

Adding energy capacity reduces the prices at which firms can sell energy since there would be more supply. Removing the grants means that existing energy generation firms need to worry less about how increased supply would harm their prices.

This mostly mimics my own experience. I’ve mostly gotten value out of handing off planned/scoped coding tasks to LLMs. It’s faster to have the LLM generate code and the quality is usually fine if the task is properly scoped.

Actually writing software was only like 15-20% of my time though so the efficiency wins from having an LLM write the code is somewhat limited. It’s still another tool that makes me more productive but I’ve not figured out a way to really multiplicatively increase my productivity.

As far as I know, by default Git doesn’t enable the “reuse recorded resolution” feature so if you made a change to the first commit you’d have to manually do the same thing for any subsequent commits.

> The CLI has nice syntax highlighting to show you the shortest valid prefix, so for example, right now I have something that looks like lzrvnkxl but the initial l is in purple, while the zrvnkxl is in grey. This means I can just use the l when referring to the change. That can be harder to demonstrate in a blog post, which can't know how to highlight this, and so often they have no highlighting.

This is such a simple UX feature that I have ended up using all the time after I switched to jj a few months back.

> we've been told to hire h1b ONLY

If you’re not being hyperbolic, this seems like fraud on the part of whoever is submitting visa applications for your company.

How far are you driving in a typical day?

If it’s less than like 100 miles (161km) I think that the vast majority of EV batteries are going to get you where you want to go, even with 25% reductions due to cold weather. FWIW, the American average is around 36 miles/day.

There is a semi famous YouTuber named Hank Green that lives in Montana and daily drives an EV. He occasionally makes videos about his experience.

According to the military release [1]:

> Approximately 700 Marines with 2nd Battalion, 7th Marines, 1st Marine Division will seamlessly integrate with the Title 10 forces under Task Force 51 who are protecting federal personnel and federal property in the greater Los Angeles area.

It seems like Trump has not invoked the insurrection act but instead it’s all under a different federal law. Steve Vladeck, a Georgetown law professor, has a write up [2] on Title 10 vs the Insurrection Act and some possible concerns. He posted this about the National guard but given the military release states they are being deployed to assist the nation guard under title 10 it still seems relevant. To quote the TL;DR of his post:

> The TL;DR here is that Trump has not (yet) invoked the Insurrection Act, which means that the 2000 additional troops that will soon be brought to bear will not be allowed to engage in ordinary law enforcement activities without violating a different law—the Posse Comitatus Act. All that these troops will be able to do is provide a form of force protection and other logistical support for ICE personnel. Whether that, in turn, leads to further escalation is the bigger issue (and, indeed, may be the very purpose of their deployment). But at least as I’m writing this, we’re not there yet.

[1] https://www.northcom.mil/Newsroom/Press-Releases/Article/421...

[2] https://www.stevevladeck.com/p/156-federalizing-the-californ...

> There is simply nothing to it

If this were true nobody would play the games? If people found no benefit or interest they wouldn’t play. Even if you think it’s literally just gambling, which I disagree with, people still find value in that even if it’s destructive.

> no depth

If this were true you couldn’t having skill based matchmaking. If there is no depth then everyone would be as good as everyone else. There would be no strategy to master.

> no cool mechanics

Subjective but I disagree.

> no story nothing

Chess has no story mode but it’s a game people still play. Why do video games need a story 100% of the time? I don’t get this critique.

> and once a player spent x amount of hours realizing that, you simply boot up 20 years old game, so you can have some dopamine rush again.

I disagree. Games today have learned from games of the past. I think most folks would find the mechanics of old games boring compared to games of today. I loved Super Mario 64 as a kid and bought it when they rereleased it for the Switch and found it so boring and infuriating (terrible controls) that I never finished the game. Same with Goldeneye — I played that game so much as a kid but there are so many better games with a similar gameplay loop today. I could probably find some fun in something like surf maps in CS1.6 but it wouldn’t hold my attention for very long.

For anyone as confused as I was, this article uses nm to mean nautical miles not nanometers.