[thisoneisreal]

I hesitate to say this because I think the AI hype is generally overdone, but I was contemplating the other day how much I would recommend my employer spend on AI tooling for me based on my salary and it's got to be in the tens of thousands of dollars per year if I'm being honest. I'm a contractor. On my most recent client, I was able to learn a completely new domain and business, design a solution and build a full stack POC (a complicated one with technologies I had never used before) nearly singlehandedly in less than 2 months. That includes all of the onboarding time, getting access to the right people, repositories, databases etc. What made it possible was 1) the AI helped me understand technologies I had never used before, from a 30 year old Java stack to graph databases, 2) it could analyze DDLs and make correct inferences about what tables and columns meant in both business and technical terms, 3) it generated mostly-correct code in multiple languages to achieve what I needed it to do, and 4) it referenced libraries I didn't know existed to accomplish tasks I needed done. I could cycle between these activities and iterate FAR more rapidly than I ever have before. To me this is the thing the AI companies should really be emphasizing. I'm not convinced autonomous agents are really working out, but I'm 1000% sold on their ability to empower developers. And I say this as a pretty skeptical late adopter who only knows the most basic AI tooling.

[skydhash]

> I'm a contractor. On my most recent client, I was able to learn a completely new domain and business, design a solution and build a full stack POC (a complicated one with technologies I had never used before) nearly singlehandedly in less than 2 months

But did you do it well? It’s not a snack, but a question based on general observation that every is praising the process, but no comment on the outcome.

[thisoneisreal]

I think I've done a good job straddling the delivery/maintainability line thus far. In the near future I'm likely going to have to downshift into smaller, higher quality changes and refactors, but I've been careful to keep an eye on the output and make sure the functions are reasonably scoped, documented, and understandable. I don't anticipate any issues maintaining what I've produced so far, but it could easily get out of hand as more concepts get mixed in.

[HDThoreaun]

Who cares? The business paid. Software engineers often overestimate how much businesses care about doing a job well vs quickly/cheaply

[skydhash]

> Who cares?

The future scapegoat, aka maintainer.

[whattheheckheck]

More agents!

[oblio]

For consulting, sure. Most of the work is hitting the ground running and figuring out the old tech stacks, creating PoCs and then moving them to production and them moving on to the next "engagement".

I think the problem is more for long lived projects and developers as regular employees. As you keep working on the product, you should be developing your own mental model of everything: business, tech stack, how the company approaches things. And that's where LLM value drops down a lot, or it should, otherwise there is a bigger problem with the developer. A huge LLM cost for a consultant is in my opinion a lot more justified than a huge recurring cost for someone who's been working on the same thing for at least 9+ months.

[fc417fc802]

Even when I know a domain well an LLM used well allows me to arrive at an answer in minutes that would otherwise have taken hours. It also accelerates onboarding an unfamiliar stack but the benefit doesn't suddenly disappear once that first hurdle is cleared.

After arriving at a seemingly workable design it helps spot oversights I've made either directly or by providing a stand in for a person to walk though the logic with (unlike a real person it doesn't become annoyed or impatient). Then it magically materializes any and all boilerplate along with 90% of the solution near instantaneously.

Yeah it confidently makes some truly bizarre errors that no half competent person ever would. Failing to go over the output with a fine toothed comb is inevitably leaving landmines in place. But I agree with you that your own mental model is required - if you have that then you will be able to spot the pitfalls assuming you actually put in the effort to read the output and really think about it instead of skimming. (As an example an LLM recently materialized an nftables filter chain for me that would have taken me quit a while to figure out on my own but for no apparent reason omitted a single key detail that would have completely broken all networking if it had been used blindly.)