991–1000 of 1001 posts
Earlier quoted context omitted.
That’s why I built all my geo infrastructure from scratch from osm, there’s still some issues but for AI location grounding it outperforms google places for $300/mo
Google has monopoly of places API, I tried foursquare and their public dataset and got tons of wrong locations of the places. It even got the location of the Eiffel tower wrong.
Earlier quoted context omitted.
Google has monopoly of places API, I tried foursquare and their public dataset and got tons of wrong locations of the places. It even got the location of the Eiffel tower wrong.
That’s why you have to make your own geocoder
Ok so a few thoughts as a former Seattleite: 1. You were a therapy session for her. Her negativity was about the layoffs. 2. FAANG companies dramatically overhired for years and are using AI as an excuse for layoffs. 3. AI scene in Seattle is pretty good, but as with everywhere else was/is a victim of the AI hype. I see estimates of the hype being dead in a year. AI won't be dead, but throwing money at the whatever U…
except a lot of people really do hate AI
Earlier quoted context omitted.
I'm experimenting with the gemini 3 and will do opus 4.5 soon, but I've seen huge jumps doing EE for construction over the last batch of models. I'm working on a harness but I think it can do some basic revit layouts with coaxing (which with a good harness should be really useful!) Let me know what you've experienced. Not many construction EE on HN.
EE on HN and from construction! I use to draft in AutoCad and Revit before switching to software. Saw your comment around using Gemini. I’d love to chat with you. I started building something for the build side of the electrical world, but eventually want to make the jump to the design side of the house.
Earlier quoted context omitted.
I hate scammers like many of the Anthropic employees that post every other week "brooo we have this model that can break out of the system bro!" "broo it's so dangerous let me tell you how dangerous it is! you don't want to get this out! we have something really dangerous internally!" Those are the worst, Dario included there btw, almost a worse grifter than Altman. The models themselves are fine except Claude that c…
Dario wishes he was the grifter Altman is. He's like a kirland brand grifter compared to Altman. Altman is a generational level talent when it comes to grifting.
Earlier quoted context omitted.
The thing is, Claude Code is great for unimportant casual projects, and genuinely very bad at working in big, complex, established projects. The latter of course being the ones most people actually work on. Well either it's bad at it, or everyone on my team is bad at prompting. Given how dedicated my boss has been to using Claude for everything for the past year and the output continuing to be garbage, though, i don'…
I was going to try having an AI agent analyze a well-established open source project. I was thinking of trying something like Bitcoin Core or an open-source JavaScript library, something that has had a lot of human eyes on it. To me, that seems like a good use case, as some of those projects can get pretty complex in what they're aiming to accomplish. Just the sheer amount of complexity involved in Bitcoin, for insta…
Earlier quoted context omitted.
I'm no apologist but this statement doesn't ring for me. It's easy to shock yourself with electricity, is it a bad tool?
Electricity isn't a tool, it's nature. An unenclosed electrical plug which you had to be really careful when handling would be a bad tool, yes. A tool is something designed by humans. We don't get to design electricity, but we do get to design the systems we put in place around it. Gravity isn't a tool, but stairs are, and there are good and bad stairs.
Of course it's bad. It's new. But it won't always be either of those things. I think "bad" is relative assessment and based on a build-up of knowledge, often over decades.
Electrical plugs and stairs are "good" only because that knowledge has been discovered and has been regulated. Expecting a tool to be literally and metaphorically fool-proof immediately upon discovery strikes me as pretty disingenuous.
In the case of AI, the most anti-AI crowd are often vehement with their fingers in their ears saying "it's not good and never will be, and shouldn't exist." To be fair, the pro-AI crowd are often raving as if all the kinks had already been worked out.
I am not in Seattle. I do work in AI but have shifted more towards infrastructure. I feel fatigued by AI. To be more precise, this fatigue includes several factors. The first one is that a lot of people around me get excited by events in the AI world that I find distracting. These might be new FOSS library releases, news announcements from the big players, new models, new papers. As one person, I can only work on 2-3…
> The fundamentals actually haven't changed that much in the last 3 years Even said fundamentals don't have much in the way to foundations. It's just brute forcing your way using a O(n^3) algorithm using a lot of data and compute.
Earlier quoted context omitted.
But Claude cannot code at all, it's gonna shit the bed and it learns only on human coders to be able to even know an example is a solution rather than a malware...
Every greenfield project uses claude code to write 90+% of code. Every YC startup for the past six months says AI writes 90+% of their code. Claude code writes 90+% of my code. That’s today. It works great. I have a faster iteration cycle. For existing large codebases, AI modifications will continue to be okay-ish. But new companies with a faster iteration cycle will outcompete olds ones, and so in the long run most…
Earlier quoted context omitted.
Well yeah. And because when an expert looks at the code chatgpt produces, the flaws are more obvious. It programs with the skill of the median programmer on GitHub. For beginners and people who do cookie cutter work, this can be incredible because it writes the same or better code they could write, fast and for free. But for experts, the code it produces is consistently worse than what we can do. At best my pride dem…
In my experience (with ChatGPT 5.1 as of late) is that the AI follows a problem->solution internal logic and doesn't think and try to structure its code. If you ask for an endpoint to a CRUD API, it'll make one. If you ask for 5, it'll repeat the same code 5 times and modify it for the use case. A dev wouldn't do this, they would try to figure out the common parts of code, pull them out into helpers, and try to make…
The broader a request is, the more likely I am to get a bunch of bloat. I think this is partly because the LLM will always try to fully solve the problem entirely from your initial prompt. Instead of stopping to clarify something, it'll just move forward with doing something that technically works. I find it's better to break things into smaller steps so that you can "intervene" if it starts to do things wrong.