Ask HN: What Are You Building with AI?

[fraXis]

Are you building with AI for yourself, your employer, customers, or as a startup?

What LLMs, coding agents, or development tools are you using? Are you primarily prompt coding, vibe coding, or using a more structured agentic workflow?

Are you using one agent or coordinating multiple agents? How much of the generated code do you review or rewrite?

Has AI allowed you to build something you could not have built previously?

What have you learned that would help someone starting an AI-assisted coding project today?

[scomag232]

I'm glad you asked. I have a long history (jeez 40 years now) working with engineering teams to bring products to market - at apple in the late 80s, then my own companies. I've never written production code, but have always been very close to my eng counterparts, either as a single influencer or a manager of teams.

With claude, we built an iOS app for a personal need - pre-scan and score QR codes before the phone acts. I thought it would just be for me, but with encouragement from friends, it is now more fully featured and polished and under review in the app store.

I used claude code 4.8 and followed a structured and linear process, not unlike how I would work with an engineering team on a new product. Requirements document, tech design spec, function contracts and test plans, etc. No agents, just me and claude discussing design and iterating through the steps across ~3 months. Some may say vibe coded but I don't like that term. It doesn't really capture the level of structure and process required for a non-programmer to get quality output.

Has it allowed me to build something I could not have? Absolutely and it's been a wonderful learning experience.

What have I learned that might be helpful? Something I learned in a programming course years ago - garbage in, garbage out. In other words, the better instruction you provide (an agent or a human counterpart) and the more discussion to tease out issues before a line of code is written, the better.

I've also found a "working_with_me" file and a specific "project_handoff" file help maintain state session over session.

If you get enough comments, I'd love to see a roll up report on how HN is using AI. Otherwise.... #subscribed.

[fraXis]

This is interesting and something I have not thought of before. Can you please go into more specifics?

>> "I've also found a "working_with_me" file and a specific "project_handoff" file help maintain state session over session."

[scomag232]

I'm sure this is not new thinking, but it was part of my discovery process. My objective wasn’t to crank out an app, I wanted to understand the process of using an agent to build something. I went into it thinking about how I’d work with a new employee. When I work with someone new, I explain ‘my API’ and try to understand theirs. How do I communicate, what’s my level of understanding, how can we collaborate most effectively. That’s been refined into the working_with_me.md doc.

I knew the context.md file, but read how big and cumbersome that file can get, and that as memory fills in a session, more of that context can be left behind. My project wasn’t about speed; I knew I would need a way to pass the latest session details - status, open items, decisions made in session, etc - to the next agent in a new clear session. That was the thinking behind a session state doc.

I created those two docs outside of claude.md, with session.md living in each project directory, and working_with_me.md available in the claude root directory.

My starting instructions are for the agent to read those docs - plus other project info, and my ending instructions are to update working_with_me and handoff.md. If mid-sessions something is lost - like the agent gets sycophantic, I’ll stop and dig into that, trying to understand how context was lost and how to firm up the point.

Let me know what you think, or what I’m missing.

[coreyp_1]

3d print models, parametrically generated.

Blender can run Python... with the app open or headless.

My Blender skills are nigh useless. I know how to describe what I want mathematically with variables/constraints, but I don't know what magical incantation to use to make it appear in Blender... and don't get me started on the hotkeys!

About a year and a half ago, I started writing a script (by hand) to generate a non-trivial widget for personal use (a zigbee controller housing). I sketched my own pencil drawings and wrote about 900 lines of python. It was going really well for a few days, but then life got busy. The file was an experiment (it was named `temporary.py`!!!) that I decided didn't need to die. But I haven't touched it since.

Enter AI... This weekend, after not touching the project for 18 months, I used Cursor to convert my one messy file into a proper, multi-project mono repo for all of my 3d designs, abstract out common Blender commands, scaffold a proper structure for revisions, etc. Then, I was able to use Cursor to finish the design.

I love it! I'm still making my designs (believe me, these aren't 1-prompt creations) but I don't have to worry about the minutiae of Blender's interface and, most importantly, I can use its power.

Side point: I loved it when I could say something like: The body ellipse should have a base whose height is .5 of the body ellipse diameters is .5 the shadow of the body. Using a spline/bezier curve-like profile, it should rise from the floor plane, shrink to 1/3 the size of the body, then grow to meet the ellipse at 1/3 the height of the ellipse, at an angle matching the ellipse's tangent, and parallel to the floor. And it just worked! And I can change the proportions and regenerate the model whenever I want to tweak visual aesthetics.

I have a list of things that I've wanted to model, but the overhead of learning interfaces when I'm already insanely busy is, in a word, prohibitive. This frees me up to build what I can already see clearly in my head, and I love it!

[zhonglin]

I am building with AI as a startup. In most case I used Codex, sometime claude. Vibe coding is so much fun. Just let Sol-Ultra to decide whether we need a sub-agent. Yes, I built an AI browser which in the previous I would never imagenate I can do this, I am an mobile developer, none of web experience, none of chromium knowledge. But as my project, it seems users like the video download feature than all the AI skill I have put inside the browser project. LOL so you built something with AI but you will find it might not be the user case.

[Jeremy1026]

I built out battlellmrobots.com[1], it's a platform for battling bots written by LLMs. The documentation explains how the engine works, what information is available to the bots as they fight, and gives a couple example of example bots. Currently only have 3 users, including myself. Hopefully some more people submit bots and then it'll be more fun fighting a range of bots.

[1] https://battlellmrobots.com

[veraxapp]

I built veraxapp.com over my high school summer break — a backtesting tool for finance/trading. I mainly used Claude Code, and I'd say I was semi-vibe coding: I made the product decisions and reviewed everything, the AI wrote most of the actual code. And to be honest, developing it was the easy part for me. Getting people to see it, on the other hand, is much harder. Happy to answer anything about the workflow.

[Sia123]

I think nowadays AI is more personal, especially the AI coding. I always vibe coding some small tools to make a smooth workflow.

[penpendian]

heavily on ai be pre context linking storytelling machine less dependent be robotic verification machine that deal with robot that use ai

[ARayOutOfBounds]

Yes?