i've stopped using google since forever now, i love how claude in terminal spoils me w websearch and stuff :P
tbh if you come across some good SEs, i'm open to suggestion. tried dia for a bit but that's a bit of an underkill
i didnt realize i needed the volume scaling with tokens burned as much as i do now xD
imagine the screaming when it confidently refactors something for 40k tokens and then finds out the thing it deleted was load bearing
the solution awareness problem is real and honestly undersold as a challenge. you spend half your energy explaining why someone should practice a skill before they're in the situation, not just selling the product itself. would love to compare notes, checking out getsimul now
this is the right frame. i'd add one layer though, the architecture itself is built from user understanding that compounds over time. every conversation with a user, every weird edge case, every "wait why did they do that" moment is load bearing. someone copying your product on day one is copying a snapshot of what you knew then, not what you know now
and then by the time they ship their copy you've already moved. the gap isn't the code it's the accumulated context about why the code is the that it is
the execution vs idea debate always felt like a false binary to me. bad idea executed perfectly is still a bad idea. but also a great idea with zero distribution instinct goes nowhere. your comment made me think about is that most founders who say "ideas don't matter" had a good idea and dont realize it because prolly they were too close to it
the begging thing is a bit of a myth imo. what it actually looks like is you describe the problem out loud to someone and they go , god yes that's exactly it - before you've even mentioned a solution. that reaction is the signal. you're not manufacturing demand you're just finding where it already exists. the hardest part is you need to talk to a lot of people before you find that reaction, that was the biggest pitfall i fell into during my first startup. trying hard not to make the same mistake twice, youve to be really mindful
the vibe coding point is the one nobody's saying out loud enough. it doesn't just make building faster, it makes the gap between "i have a thing" and "i have a business" feel smaller than it actually is. you ship in a week and it works and you think you're close. you're not, you just moved faster to the same hard problem
the distribution bit is where we are right now honestly. talking to users before writing most of the code was the one thing that changed how we think about it
the feedback loop point is exactly it. flight simulators work because every decision gets immediate feedback. most real life skill practice has the worst feedback loop of all, sometimes you don’t find out you negotiated badly until years later. that’s what we’re trying to compress
the thing that gets me every reread is the structure of the joke. same question, asked across the entire lifespan of the universe, same answer every time. asimov could have made it tragic but instead it reads almost like a bit that keeps escalating and then the punchline is that the answer was always going to come, just on a timeline so absurd it laps back around to funny
the LLM parallel does hit different on this read multivac says insufficient data across ten trillion years and the whole story is basically if more compute and more data eventually gets you there. what's weird is the story answers yes, not on any timeframe that helps the people asking tho.
feels uncomfortably close to the actual situation where the models keep getting better and the answer keeps being "not yet, ask again later" while the answer is getting ready years late
ted chiang if you haven't already. story of your life, exhalation, the lifecycle of software objects. same thing asimov does where the sci fi premise is really just a frame for a very human question. except chiang does it in like 30 pages and you feel it for a week
literally put up a link get community feedback, product hunt, hackernews, reddit, linkedin, as many as you can - talk to your users, see what's been up. my personal go to is always user-first approach if you're not building for the user they wont use it. core mantra. talk and iterate make them feel they're a part of the growth
oh my god, this is gorg. i love museums for the same exact thing. there's so much you donno and every visit just leaves me in awe. thank you for sharing it. big wide bful world
slow was the safety net for sure but then there were errors too, there's a sweet equilibrium spot where ai + human oversight reaches that efficient + almost perfect situation. ofc with the right methods
this is more on the lines of the marketing play that "developers are out of job" or "oh i created an app overnight", well sure you did but you also put yourself at a lot of risk, a lot of ai tools are just so risky at times, i'm tbh surprised why claude still asks for api keys to be provided in terminal. you could almost bet that the devs building it 100% know they should not do it, but there's always a way you can brainstorm with the ai, crosscheck with stackoverflow, reddit or anything like you did earlier. this is no substitute for the way programming worked, just a 100% faster and efficient engine for sure. it's such an under-explored area that there's def more eyes needed on it
i mean i don't disagree with you, but then there's at least unit test cases that we write as devs in a certain way that ai just doesn't understand quite well yet. we worked so much on speed that we forgot along the way that slow was a safety net. we tend to let our qa teams leverage ai as much as they want to automate test cases but i don't feel super comfortable getting rid of it just yet
Every other field that's figured out high stakes failure models eventually landed on the same solution - make sure two people that understand the details are looking at it - pilots have copilots surgeons with checklists and nuclear plants have independent verification. Software was always the exception, cause when it broke it mostly just broke for you, vibe coding is not going to change the equation, it barely removes one check that existed before is that the people who wrote the code understood what was going on, but now that's gone too
interesting, so the ai got the hard stuff right. password hashing, schema design, fine. it fumbled the stuff that isn't really "coding" knowledge, feels more like an operational intuition? backup folder sitting in web root isn't a security question, it's a "have you ever been burned before" question, and surgeon hadn't. so they didn't ask and the model didn't cover it, imo that's the actual pattern. the model secures exactly what you ask about and has no way of knowing what you didn't think to ask. an experienced dev brings a whole graveyard of past mistakes into every project. vibe coders bring the prompt