iagooar

7 ms·

91.
by iagooar 1y ago
This issue has been one of the biggest issues with the Claude models, not so much with GPT-4 or GPT-5. I even had this Cursor rule when I was using Claude: "- Do not use statements to catch all possible errors to mask an error - let it
92.
by iagooar 1y ago
Wow, half of this article deeply resonates with what I am working on. Text-to-SQL is the funniest example. It seems to be the "hello world" of agentic use in enterprise environments. It looks so easy, so clear, so straight-forward
93.
by iagooar 1y ago
As a non-native English speaker, I will not be mistaken as AI for a while. Occasionally, I will introduce a small typo just to make sure people don;t think this is AI.
94.
by iagooar 1y ago
Oh and I agree so much. I just shared a quick first observation in a real-world testing scenario (BTW re-ran Sonnet 4.5 with the same prompt, not much changed). I just keep seeing how LLM providers keep optimizing for benchmarks, but then I
95.
by iagooar 1y ago
Yes, I did. I ran the test again, took Claude ~4mins this time. There was no error now with the auth, but the functionality was totally broken. It could not even find the most basic stuff that matches perfectly.
96.
by iagooar 1y ago
I even added a disclaimer "anecdotal evidence". Believe me, I am not the biggest fan of Sam. I just happen to like the best tools available, have used most of the large models and always choose the one that works best - for me.
97.
by iagooar 1y ago
I think that is an interesting observation and I generally agree. Your point about prompting quality is very valid and for larger features I always use PRDs that are 5-20x the prompt. The thing is my "experiment" is one that repre
98.
by iagooar 1y ago
Anecdotal evidence. I have a fairly large web application with ~200k LoC. Gave the same prompt to Sonnet 4.5 (Claude Code) and GPT-5-Codex (Codex CLI). "implement a fuzzy search for conversations and reports either when selecting &quot
99.
by iagooar 1y ago
I would like to share an even more extrem version of this. I come from a country that could be potentially affected by the Russian-Ukranian war. A couple of years ago, the government presented a program for volunteers, consisting of a milit
100.
by iagooar 1y ago
I've had whale when I went to Norway. Was it good? It was OK. Would I have it again? Probably not. Still a nice experience. Also trying cured reindeer, moose and seal slices was pretty unique.
101.
by iagooar 1y ago
Finnmark in Northern Norway is more like 1.5-2, Troms is 3-6. But I agree, Alaska and Western Canada are places to go some day. Funnily, as a student I had this offer to go to work as a waiter on Alaska. For different reasons that didn&#x27
102.
by iagooar 1y ago
I have been to northern Norway and Iceland a few times. It all started when I was running my previous SaaS at the peak of its growth. I needed a break, and wanted to go far away, while still being close enough if sh*t hit the fan. My co-fou
103.
by iagooar 1y ago
I know. But why keep distributing over npm?
104.
by iagooar 1y ago
Even OpenAI uses npm to distribute their Codex CLI tool, which is built in Rust. Which is absurd to me, but I guess the alternatives are less convenient.
105.
by iagooar 1y ago
High crime, high taxes, stagnating economy. I am surprised Germany landed this high on the list.
106.
by iagooar 1y ago
Book a demo! Really, it will not be self-service just yet, because it requires a bit of holding hands in the beginning. But I am working on making a solid self-service signup experience - might need a couple of weeks to get it done.
107.
by iagooar 1y ago
200k LoC + 80k LoC for tests. I have roughly 2k tests now, but should probably spend a couple of days before production release to double that.
108.
by iagooar 1y ago
I add the MCP server ( https://ui.shadcn.com/docs/mcp ) Then I instruct the coding agent to use shadcn / choose the right component from shadcn component registry The MCP server has a search / discovery tool, a
109.
by iagooar 1y ago
It is more than 200k lines of slop. 200k lines of code slop, and 80k lines of test slop.
110.
by iagooar 1y ago
Fair enough. I spend entire days working on the product, but obviously there are lots of times I am not running Codex - when reviewing PRDs, testing, talking to users, even posting on HN is good for the quota ;)
111.
by iagooar 1y ago
Then I will adapt and expand. Have done it before. I am not giving universal solutions. I am sharing MY solution.
112.
by iagooar 1y ago
Yes, I have over 2 decades of programming experience, 15 years working professionally. With my co-founder we built an entire B2B SaaS, coding everything from scratch, did product, support, marketing, sales... Now I am building something new
113.
by iagooar 1y ago
I 100% agree with the over-fitting part. I tend to provide detailed PRDs, because even if the first couple of iterations of the coding agent are not perfect, it tends to be easier to get there (as opposed to having a vague prompt and move o
114.
by iagooar 1y ago
I added that line, because otherwise the LLM would generate goals that are not verifiable in development (e.g. certain pages to render <300ms - this is not something you can test on your local machine). Documentation is a different topic
115.
by iagooar 1y ago
Tools change, standards do not. My process is probably more robust than simply reviewing each line of code. But hey, I am not against doing it, if that is your policy. I had worked the old-fashioned way for over 15 years, I know exactly wha
116.
by iagooar 1y ago
Strong feelings are fair, but the architect analogy cuts the other way. Architects and civil engineers do not eyeball every rebar or hand compute every load. They probably use way more automation than you would think. I do not claim this is
117.
by iagooar 1y ago
Codex uses this principle - /review runs in a subthread, does not see previous context, only git diff. This is what I am using. Or I open Cursor to review code written by GPT-5 using Sonnet.
118.
by iagooar 1y ago
Telling which component to use works perfectly too, if you want a very specific look.
119.
by iagooar 1y ago
A small description can be extrapolated to a large feature, but then you have to accept the AI filling in the gaps. Sometimes that is cool, often times it misses the mark. I do not always record that much, but if I have a vague idea that I
120.
by iagooar 1y ago
It is true AI-generated UIs tend to be... Weird. In weird ways. Sometimes they are consistent and work as intended, but often times they reveal weird behaviors. Or at least this was true until recently. GPT-5 is consistently delivering more

More ›