◼ Software engineer & builder
◼ 15+ years turning vague ideas → products
◼ Building with AI, still reading diffs, sharing what works
◼ Co-founder of slay_lines
alright X, here we go again
I got bored after burning through my OpenAI tokens and made an English account
I quit my job to focus on building, and I think I have something to say about tech
I'll keep it personal without pretending I have it all figured out
@ZappoMan@arvidkahl I still use them extensively myself. AI tends to write shallow tests that mirror the implementation. For better coverage, you end up writing tons of them. It’s much harder for AI to do that with behavioral integration tests. I believe that’s why there is that chatter.
@frankdejonge here's the neat part: we don't
/jk I actually think shared activities and hobbies that don't feel obligatory or create any pressure are great. but internally, you still need to prioritize them over something else in your life, and that almost never happens
@din0s_@JFPuget good stuff. related to this, I've been thinking about the term "cognitive debt", and I'd say it's acceptable when it's fully acknowledged and tracked just like technical debt
@JFPuget the worst thing is that claiming something is done doesn't mean any actual useful work was done anymore. verification and ownership still cost time, but you can't tell if that time was well spent without verifying it yourself
@mitsuhiko I used to think it's ugly but lately I've been liking rust code too. I'm now wondering what kind of brain damage caused the change in attitude lol
actually, it made me think that sometimes designing while developing is the only way to figure things out, so I wouldn't call it inferior, but if you find yourself doing everything this way, you're probably moving too slow. I learned it the hard way.
@thorstenball it sounds more like a planning skill than blind shipping
shipping well-planned slices of work >> shipping one big chunk of work that's poorly planned so you end up making design decisions as you go >> shipping chaotic, unsystematic small changes
@thorstenball it sounds more like a planning skill than blind shipping
shipping well-planned slices of work >> shipping one big chunk of work that's poorly planned so you end up making design decisions as you go >> shipping chaotic, unsystematic small changes
@thorstenball it can also test some obscure scenarios like slow internet or a dropped connection. I’m now considering making playwright e2e tests a priority in every web project and focusing heavily on their reliability and coverage
@nlytheloserswin@AndrewYNg true, but I find targeted attacks far more dangerous than unintended damage
I agree with Andrew that this risk can be mitigated by restricting access and putting stronger safety controls in place
@EmileAndHisBots@AndrewYNg this fear is real too, but not as much at this point in history. machines aren’t that close to having this capability yet imo