Similarly, I used gen ai to review a real estate purchase. I provided Zillow listing photos and serial numbers of all appliances, the electric panel, and a few additional not pictured areas that I took during the walk through.

I prompted the AI to write a report as if it were a home inspector and it actually did a better job and identified some issues the paid 750 usd inspector missed.

From pictures alone? What are some examples?

I presume something like this: https://www.penny-arcade.com/comic/2007/06/22/perfectly-reas...

It noticed a flooding area due to low grass by the walkout door. It noticed mixed 15 and 20a receptacles on the same circuit. It noticed warped siding and recalled circuit breakers still in use.

15A and 20A receptacles on the same circuit sounds fine as long as it's a 20A circuit? And how could it tell which outlet is on which circuit?

It can’t, but it’s read reports before so it sure can simulate an answer.

To give it the benefit of doubt, it's possible it saw a circuit labelled "kitchen" in the panel, and then in photos of the kitchen saw mixed outlets.

(I'm not in the US - would a 'home inspector' actually go around buzzing out outlets anyway?)

They won't necessarily map out all the circuits but they will generally test them all with a tester to find wiring problems.

Yes, most will at least test GFCI receptacles especially in the kitchen. I bought one to test my basement after a renovation.

What, the Zillow listing of you home doesn't have pictures of mixed 15 and 20a receptacles on the same circuit that an AI caught but that an inspector missed?

Is that what you're telling us??

It is useful if you automate generating release notes. Then your notes are grouped by new features first, then bug fixes after. This makes it a little easier for non-technical uses to read.

Commit messages are good release notes rarely.

it's usually a "something is better than nothing" situation.

If you have somebody willing to write custom release messages, that's definitely better; but conventional commits is better than nothing for it.

Absolutely not. Commit messages should never be automatically passed through to the end-customer. I also worked in a place that tried it once and it was a disaster. Sure, a list of commit messages can be a useful start as a list of things that might want to be put in the release notes, but very rarely is the developer the right person to be explaining those changes to the end user.

If a developer is being asked to do that, it's a good sign that the PM isn't doing their job properly.

They did not say generated release notes are useful if you care so little you would write no release notes without them however.

That's right, but with AI help + some hallucination you can get nice looking release notes out of the worst mess of commits.

You can have a writer re-write them into acceptable release notes. It gives them a good and accurate starting point.

Closed issues are a better starting point in my experience.

I use Optery for about two months a year, seems to do a good enough job for most of the data brokers. There are also discounts or promo codes to lower the price as well.

https://www.optery.com/

HN Launch: https://news.ycombinator.com/item?id=30605010

Promo codes: https://www.optery.com/optery-promo-codes/

I tried Optery. It got a good chunk done in two months, then the rest were just pending for a year... until I cancelled. Felt like they were just keeping me on the monthly dole while they didn't do anything.

The other advantage with bash is that most developers can run it locally to validate what it is doing and debug issues. With GitHub Actions you need to always commit and push, slowing down the DX.

Shameless plug: solving this "push and pray" problem is something we have been focusing on with Dagger. It's an open-source CI platform that decouples the runtime from the triggers. The runtime is open source and local-first, so you develop the actual logic of your pipelines with a proper dev loop. Then, you separately wire up your git triggers. The same pipeline logic can be triggered locally or from git events.

IMO this is the only clean way to solve the problem. If you want to check it out and share feedback: https://dagger.io . We also have a very active Discord server full of CI nerds.

You should add an Easter egg in your cli program: dagger attack, which prints out a favourite Top Gun quote.

Yep, I tried to use Act to get a sense of what our YAML was doing but it failed to pull the docker images and I gave up - not enough incentive to test locally when I can push to GH and yolo it and hope the ops folks can help me figure it out

Commit, Push, & Pray.

LMS’s are a lot like programming languages. There’s the ones people complain about and the ones no one uses.

I'm an LMS admin and yeah, that sounds about right.

It depends on what you pay for. If you need FedRamp or IL4+ compliance you are likely on dedicated infrastructure. Everyone else uses multi tenancy.

VSCode has this too: chat.disableAIFeatures

Sources:

https://news.ycombinator.com/item?id=46501220

https://code.visualstudio.com/updates/v1_104#_hide-and-disab...

I wonder what will happen to the entire legal system. It used to be fairly difficult to create convincing photos and videos.

AI can probably fool most court judges now. Or the defense can refute legitimate evidence by saying “it’s AI / false”. How would that be refuted?

For better or worse, the only admissible evidence going forward will probably be either completely physical or originated in attestation-capable recording devices, i.e. something like a "forensics grade" camera with a signing key in trusted hardware issued by somebody deemed trustworthy.

Given the obvious personal safety upsell ("our phone/dashcam/... produces court-admissible evidence!"), I think we'll even see this in consumer devices before too long.

By having people also testify to authenticity and coming down like the hand of God on fakers, the same way we make sure evidence is real now.

Yes, that is a major worry of mine, too. CCTV evidence is worth nil now (could be generated in whole or part), and even eye-witness testimony can be trusted (sure, a witness may think they saw the alleged perpetrator, but perhaps they just saw an AI-generated video/projection of someone).

Trials have rules for evidence. You can't just pull out some footage out of nowhere. Where did that come from? From what camera? What was the chain of custody on its footage? Etc.

If it means anything, I have a 1990 Almanac from an old encyclopedia that warns the exact same thing about digital photo manipulation. I don't think it really matters at this point

MS13 was literally tattooed on his knuckles!

Multiple data sources, considering the trustworthiness of the source of the information, and accountability for lying.

You might generate an AI video of me committing a crime, But the CCTV on the street didn't show it happening and my phone cell tower logs show I was at home. For the legal system I don't think this is going to be the biggest problem. It's going to be social media that is hit hardest when a fake video can go viral far faster than fact checking can keep up.

Say what you about the Anna’s Archive Spotify scrape: it made me realize how much music exists and how much music was never listened to.

If every track was 3 minutes long, it would be about 1450 years worth of music. You can never experience it all.

You could if you parallelized the operation. Probably tantamount to torture though.

We're shipping a skill file with the CLI: https://skills.sh/github/gh-stack/gh-stack

Everyone will have their own way of structuring stacks, but I've found it great for the agent to plan a stack structure that mirrors the work to be done.