Hacker News Reader

Comment by yuppie_scum | Hacker News Reader

Quality just isn’t there with offshoring due to the time zone difference and distance. Accountability isn’t there. The development savings will be offset by onshore QA, PM, Project management and rework by onshore devs. Been seeing that play out again and again for 20 years.

I have seen slightly better results with offshoring from Latin America - fast feedback makes a huge difference - being able to meet frequently and provide feedback in the same time zone makes a huge huge difference.

I think that AI will rapidly slot into this niche of “indirect programming” however. As soon as token sizes allow large codebases to be be parsed and modified, it’s all over.

View on HN

> I think that AI will rapidly slot into this niche of “indirect programming” however. As soon as token sizes allow large codebases to be be parsed and modified, it’s all over.

Context size is the least of an issue - current models suck in capability.

I've been building stuff on top of openai API for months and getting it to do trivial stuff with instructions reliabily is incredibly hard. These are things like "if a user asks a question about an individual but only specifies first name ask for last name before invoking a search function" - even with examples of what to do you'll get GPT 4 ignoring instructions and calling search with just a first name.

Like I understand that the potential they are showcasing is a breakthrough, and people are right to be excited about that - but we are extrapolating way to much based on breakthrough improvement pace.

We've seen similar claims about self driving cars being a year away since 2016.

Unless there's some breakthrough, I don't see AI affecting developer demand in the next two years at least.

> Quality just isn’t there with offshoring due to the time zone difference and distance. Accountability isn’t there. The development savings will be offset by onshore QA, PM, Project management and rework by onshore devs. Been seeing that play out again and again for 20 years.

Yupp, it pretty much turns a quick conversation at standup or a quick huddle about a change into a 16+ hour turn around time at minimum.

Also, these companies tend to be lead by engineers. They identify with the engineers and believe that this work (and thus themselves) is special. Not the case with say auto manufacturers.

Who is to say the future will "remain this way" for engineers? When will it become more of a commodity?

More and more these companies are lead by groups of people who have never worked as an engineer.

I think most CTOs of smaller companies make $500k/yr or more, don't write code, and are just really good at driving delivery/working with others/having good instincts on trusting others (like enterprise architects, dev leads, etc.)

Isn't the current AI level already good enough to offload a big chunck of the work? I feel like the technology is already there and it's just a matter of adoption

Reply on news.ycombinator.com