bentcorner

I will always have a soft spot for the original Prince of Persia. It was one of those games I played constantly as a child, although only when my dad would let me use his Apple ][c.

I only realize it now but it had some very unique game mechanics that even today you don't see very often (ok maybe that's a bit of a stretch but the mechanics were novel to me back then):

- Notably you have 60 minutes to finish the game. Dying doesn't reset the timer, so there is constant pressure to keep moving.

- There is a satisfying parry mechanic. This is still rare to see in 2d platformers.

- Incredibly smooth animation. This could be nostalgia goggles but the rotoscoped animation really stood out compared to other games of the era.

> Watching it do log file analysis in seconds that would have taken me hours (edit: days really), and which I would therefore never have done in the first place.

Just today I had my agent diff two logs to find a very nitpicky difference that was the cause of a problem, I pointed it at a ADO extension that was having issues, it downloaded the VSIX and decompiled the .NET binary to verify. Based on that information it suggested a workaround which I was very skeptical of, but well it worked.

All of this I technically could have done but I probably wouldn't because it would have taken too long without a clear payoff.

Anecdotally Windows ARM works fine for me, although to be honest most of my work is command line + browser anyway. WSL works like a treat. Steam installs and most lower end games also play fine on my ARM laptop too. Games that require kernel anticheat don't work.

I think they make a great "second device" where you have something meatier to fall back to if something doesn't quite work right. I'm not sure if it's ready to take on the "main device" role just yet. But it's a far far better experience than the Surface RT days.

I'm trying not to spoil anything but this article reminds of this excellent short story: https://asteriskmag.com/issues/07/sins-of-the-children

> Also I just find it a little insulting if someone sends me an AI response

I run into this with AI-generated PR comments. I think where I work we are still grappling with the "right point", because LLMs can certainly provide valuable feedback, but they are not at the point where they can do so unsupervised, and to do so just feels unprofessional.

And there's another layer where it is even worse when a colleague spends the time critiquing code and someone (or something) replies to the comment with mostly useless filler. It's like being handed a small hand-crafted gift and then throwing it into the garbage in front of them.

I've come to the opinion that conflicts should be committed and merge fixes should be in another commit afterwards. Arguably even if the merge fix is trivial.

I'm certain there are a non-zero number of TVs that either attempt to auto-join popular wifi hotspots (xfinity/tmobile/starbucks/etc.) and/or have cellular connections for telemetry.

Thinking more on this I think a business opportunity in the future will be companies that design hardware stacks that can go in random appliances that can gather usage information in the name of telemetry.

I give it +/- 5 years before an OTS coffee maker at walmart phones home.

Anecdotally from people I know from MS is that adhoc slack usage was popular up until lockdown, at which point more resources was poured into Teams due to everything going remote and internal slacks were frowned upon.

I see a parallel here where a competitor's product is taking over and MS leaders see it becoming an existential problem and are putting their foot down and pushing internal users to the company's products.

Also if half the company don't even want to use something made by the other half it's a bad look lol

20 days is rookie numbers. I can get a million engineers to each review a single line and finish code reviewing the entire code base in a minute.

The premise of this article is incorrect - MS isn't cancelling Claude code internal usage because of AI costs too much, they're cancelling it because GitHub copilot is the compete product and they want their employees to use their product.

It's the same reason Teams got so much attention during lockdown.

> The first half is a constant flood of footage from the iPhone, the DJI Pocket, the drone, the Nikon Z8, and lately the Ray-Ban Metas too. There's always something being recorded. Every photographer or videographer I know is sitting on the same problem: an archive that grows faster than they can edit it. The second half is why mine never gets touched.

This is your second paragraph but reads awkwardly. You mention two halves in the previous paragraph, so I kind of try to map those two halves to the halves in this paragrpah. But I don't understand what the second half is in this second paragraph.

> Three months ago the lodge's social channels went dark. Not for lack of content; the lodge has years of raw footage across multiple SSDs. The bottleneck was editing time, and my time disappeared. Claude Code with Opus 4.5 (and then 4.6) hit the point in February where you could leave agents running for hours and come back to merged PRs. KaribuKit was going live with its first paying property in the same window. I stopped sleeping properly, started running three or four agents in parallel in the background, and the months when I would have cut reels turned into months when I shipped software instead.

I don't fully understand this paragraph either. Your time disappeared? Into what? Was it the lack of sleep? I don't know what KaribuKit is.

> I asked it out loud: how does the agent know what's in each clip?

Did you? Really?

> Four bugs, four lessons

I've noticed that AI tends to rathole into random things when summarizing a piece of work, so I'm skeptical that these were actually the most four interesting bugs you could have shared.

I would recommend you just remove this section or take the time to actually think about some learnings you had from this project. Syntax errors or missed CLI params are mildly interesting but what makes these four bugs interesting to your readers?

> The actual take

The same criticism here applies. Are these your real takes, or did Claude make these up too?

Some obvious tells to me of things that AI likes to write that humans rarely ever say:

> Both real, both consuming attention.

> Four constraints set the shape:

There's way more than just this (the writing style of nearly the entire post screams Opus 4.7), but that's just what jumped out at me when I started reading your post.

I don't mind you used AI to write this but in the future when you write using AI, take the time to read the entirety of the article and consider the goals of what you want to write and if the AI achieved that. Take out what doesn't belong and make sure that what you have left says things in your voice.

I maintain a part of my team's CD process and I've observed a 30% increase in PR velocity since we started adopting agentic tools but it was a "one-off" increase (as-in, it hasn't continued to increase beyond that since about a half-year ago).

I'm guessing though that there are other improvements in code quality and feature velocity. I've noticed personally that AI is really good at catching smaller things that are easy to miss (e.g., if you ask it to rename fooTheBars it also updates all the relevant comments or enums that you might have missed).

I suspect at some point AI-written code will be eventually artifacts generated build-to-build. The design docs and UI tests are the source and the model follows instructions to generate the product. If you make the models deterministic then model improvements give you code improvements across your entire codebase "for free".

I personally believe that the legacy we send to the stars will be silicon.

Robots have landed on Mars. Maybe they will even figure out how to use minerals on Mars to build more of themselves. It is plausible to me that as far as space exploration is concerned that it will be autonomous within a few hundred years.

My local school district somehow has <schooldistrict>.us

Not sure how that came about.

This feels like it deserves to live somewhere on a blog, not as a comment on some forum. This is really interesting thanks for sharing.

30% is a surprising metric because just anecdotally my team's PR submissions have jumped by about the same ratio in the same time frame. (No, I'm not measuring it - I just own a downstream piece and have casually observed its higher rate of use).

Kinda sorta, but there's a limit where users will typically install X apps and apps of Y size need Z extra space to update. User content would fill up the rest. I would imagine a typical 256 gb phone is probably over this limit and people who take lots of videos/photos just need to clean up their phone a little more often.

Exactly - I'm thinking the bad spatial navigators have a higher probability of washing out of driving and pursue some other career. They may not say "I'm bad at figuring out where I am", but the economics of the job are just a little bit worse for these people.

I don't think this is where the problem lies. If you kill someone with intent, it's murder. But the whole system needs to prove that you killed someone with intent beyond a reasonable doubt, and a DSL will not help you there.

This could in theory already happen without any tech, but I suspect since the government is pretty monolithic, any changes in a specific law are all being done by the same set of people.

You might not have merge conflicts but I imagine you could end up with conflicting guidance from two separate pieces of law (e.g., law A says you must wear green on St. Patrick's day, law B outlaws green pajamas).

I think the right solution is to endow the LLM with just enough permissions to do whatever it was meant to do in the first place.

In the customer service case, it has read access to the customer data who is calling, read access to support docs, write access to creating a ticket, and maybe write access to that customer's account within reason. Nothing else. It cannot search the internet, it cannot run a shell, nothing else whatsoever.

You treat it like you would an entry level person who just started - there is no reason to give the new hire the capability to SMS the entire customer base.

> People are free and probably do this because it is slow. Alternatives often are not a bad thing.

Alternatives are always good but IMO brew is just not something I interact with all that much and to me it's "good enough". It works and does what I expect, although to be fair maybe I'm on the happy path <shrug>.

Tik Tok is arguably a Facebook killer.

Roblox is in some ways there, I think Epic thought fortnite could have competed. IMO they made a strategic mistake in shackling their game-as-a-platform to Fortnite. I thought the music fortnite thing looked interesting, but I have negative interest in installing Fortnite.

Call it something else and make it literally the first thing you see on epicgames.com, have it work on mobile, and maybe things would be different today.

(Aside: Roblox wins because I can go from typing in roblox.com into my browser and be playing a game with a friend in under 20s)

Yeah this one is a classic: https://youtu.be/8OzZxjqKG10

Makes me wonder if there's a bet you can take on Polymarket that Polymarket will get shut down due to it negatively influencing behavior. The insider trading on that one should get interesting.

I just say something like "spawn an agent to review your plan" or something to that effect. "Red/green TDD" is apparently the nomenclature: https://simonwillison.net/guides/agentic-engineering-pattern...

I've also found it to be better to ask the LLM to come up with several ideas and then spawn additional agents to evaluate each approach individually.

I think the general problem is that context cuts both ways, and the LLM has no idea what is "important". It's easier to make sure your context doesn't contain pink elephants than it is to tell it to forget about the pink elephants.

Unfortunately the only way this changes is if a company writes a just enough unreasonable ToS, and someone violates it in just the right way and the company decides to enforce said ToS, and the user fights back, and this all ends in court.

I'd be surprised if all those stars align anytime soon.

Installing humanoid robots in a factory is like using regexes to parse data.

It makes sense if it's a one-off but there are better solutions.

Maybe it does make sense for small scale businesses that need just a little automation? Like a humanoid robot could restock shelves and do inventory in a grocery store at night, and you wouldn't need to retrofit anything to be able to do that.

Large scale factories seems like the wrong use case for humanoid robots.

Personally I stopped using Facebook because even in the before-AI days it started becoming a glamour photo book of everyone you ever knew (and probably lots of people you only kind of sorta know), and while people certainly deserve to do and see great things, seeing it all shoved in your face every day becomes exhausting in a keeping-up-with-the-joneses kind of way.

I totally get that not everybody is like that, but I am, and so I stopped going to Facebook.

These days I'm in private Whatsapp groups for my direct family and so I learn about what they do, and not the random stuff that my neighbors and 20-years-past classmates did.

My wife is still active on Facebook and I actually do still visit occasionally to boost her posts but that's about it.