Ask HN: What was your "oh shit" moment with GenAI?

1. When ChatGPT came to the masses, it trivially solved my standard phone interview problem for new SWE hires. It's not particularly complicated, but it screened out a lot of candidates pre-AI and was a good filter.

2. At one point, there was a bug in some client software we ship that was erroneously displaying a protocol-level disconnect message to clients when the server shut the connection unexpectedly.

In very few turns, ChatGPT gave me working code to intercept the error at the client level via a pseudo-server that intercepted the requests and implemented the proper teardown procedure. It essentially implemented a micro-server for the protocol we were working in as a bug workaround.

3. I'm working on a major rearchitecture (6+ month project of very senior engineers) of an internal system for scalability and maintainability with tight latency bounds where correctness with the old system is necessary.

I came back from an international vacation quite jet lagged and was having trouble doing actual coding work for a bit since my focus was shot. Taking additional days off wouldn't have helped recover from the sleep issue more quickly due to childcare responsibilities.

I tried instead building a pipeline where I'd run an integration test suite, throw a list of failing tests at Claude, and make a PR if it made more tests pass. I automated as much of the AI "loop" as possible in my state at the time.

Using this technique, a coworker and I fixed over 200 failing integration tests for parity with the old code in 1 week. We subdivided failing tests between engineers and re-sub-divided with each success.

I re-reviewed the code before we launched when I was more well-rested, but this was an extremely effective technique and makes me think that AI-enhanced test driven development (TDD) is the future.

I was shocked when it built a usable version in one shot and instead of having a prototype to demo, we basically had the whole thing done in a few hours.

Step 1, running my tests on Gemini. Having it argue two sides of radical social systems and realistic seeming implementation routes. Collaborative story telling into the absurd. Having it solve impossible seeming riddles (though that still leads to hallucinations).

2. Gemini explaining arcane BIOS settings to me not found anywhere on Google, mostly correctly.

3. Claude dissecting a tongue-in-cheek theory/blog post of mine, deeply analyzing it while catching flaws and catching on to irony and sarcasm.

4. Several non-coder friends building fully automated AI slob distribution and sales/BI platforms with Claude, and forking and greatly improving projects on GitHub. I did some tests with Vibe Coding myself, had Claude write a small game from one prompt. This is a bit insane, I must admit.

I'm a former skeptic who was written a lot on AI and society, published some, and held public discussions with experts.

I have since sung the praise of, especially, Claude in closed forums for hundreds of county and government digitalisation and security people in a way I'd have never expected two years ago.

Background, I'm an IT and security guy myself with 30 plus years of light coding and heavy, broad enterprise stuff. I'm on record now for saying things on AI I would have had myself hospitalized for two years ago.

Many of the people around me go "eh, just check your code for security" and "I tried coding in ChatGPT, i had it output some puny script, it's not that great".

In return I had Claude (free tier) make a PDF for them on uses, chances, risks, legal framework and integration with other AI and services for them, nicely formated, from one prompt. It seems 100 percent factually correct and Claude fixed a bug in the PDF generation code it pulled off of the web from the same prompt to complete the task.

I also explained to them why especially the library pool for Python and JavaScript enables Claude to write stuff that is rather impressive, and that while not fully scalable (yet), it might be "good enough" for 90 percent of the tasks people want done these days.

I suddenly find myself berating people on government forums, where some people from the national IT security advisory body are part of the (mostly silent) audience, to wake up in some way and consider what it means. Yet our national approach to security is mostly still "Vibe coding is a toy and a party trick, that will blow over eventually, get a code audit or follow some security check lists."

Meanwhile I believe that in two years, most of GitHub will be written by non coding hobbyists like my friends.

And having seen how insanely laissez faire some commercial software folks treat security and product service for extremely pricy products, maybe DIY vibe coded stuff does not even look bad in comparison.

While some of 1000s of people around me doing the "real world" work around me can't wait to give agency to the AI, many others here, including many people in charge, still seem blissfully ignorant. Or want to be.

Ultimately, I only know one thing for certain: Society is currently heading full yolo into this, and systemically, that might be the only way because that is the true nature of AI.

They say people don't understand the exponential function. This is true.

But most people also don't seem to understand the implications of this new approach to exponential pattern recognition and reassembly, that operates on the microsecond scale, and with an insane pool of information at its disposal. It will probably never make NEW stuff, but it will assemble old stuff so fast and complete it will still seem indistinguishable from magic.

Society is in for a wild ride.