One search across memories, receipts and files is a smart feature. Filtering by type should make it quick to find things.

#connect 👋

#connect 👋

Sharing the broken UTM tracking right next to the 25K-view win is what makes build in public worth reading. Respect to Saurav for posting both.

#connect 👋

Prompts and versions scattered across chats is a real pain. Keeping images, prompts and versions together in one Mac app sounds really useful. Good luck with GenTidy!

A MacBook you can take apart in the browser is such a fun idea, and making the code public is a nice touch. Great Three.js work.

#connect 👋

A plain folder works fine for storage since they are just text files, and Obsidian can open that same folder as a vault. The harder part seems to be handing them to people who don't read markdown. When you convert to Notion or Word, what do you end up fixing by hand?

The same lesson shows up on the paste side: a real heading level survives a paste or conversion far better than bold text dressed up as one. In your own docs, what breaks first when they get converted, headings or tables?

Path remapping is the part that makes this useful, since moving a session between machines usually breaks on absolute paths first. Does it handle Windows paths too? Most of my agent work runs on a Windows laptop, and that's where tools like this tend to trip.

Loops beat prompts, but a loop is only as good as its exit check. Without a hard pass or fail at the end, the agent keeps going and calls it done. I'd want to see how Loop Engineering handles a run that never converges before wiring it into anything.

This is the step most agent workflows skip. A green build says nothing about whether a real person can get through signup. Having the agent click through the main journeys and screenshot failures turns review into checking evidence, not trusting a summary.

Tools for the trades is a lane I care about. I ran heavy equipment for years before I wrote any software, and the trades are full of daily chores software barely touches. These days I build AI tools solo and test open source dev tools on Windows.

Solo builder here, making AI tools in public. The part nobody warned me about: writing code got easy, and keeping the agents consistent from one session to the next became the real work. Good luck with the SaaS. What problem is it going after?

Same, learning as I go. I spent years running heavy equipment before any of this, so my stack is Cursor, small tasks and testing things myself before I trust them. Each week I also try a few open source dev tools on Windows and post what holds up.

AI and SaaS side here. I'm a one-man shop that leans on a few AI agents, each with one job, and most of my effort goes into the rules that keep them from wandering off. Curious how much of your outbound you let AI write versus touch yourself.

Every week we test open source tools for Devs Helping Devs, and the catches teach me as much as the wins. So I'm curious: what's a tool you quit using this year, and what made you drop it?

VOX for dictation is a great idea. the catch with an always-open mic is it'll type whatever the room says too. what I went with in MARNE: it only listens when you tell it to, no hotkey needed (you can program one if you want), and Escape bails if it caught the wrong thing

Cost splitting is the right pain to go after, that's where group trips get awkward. Good pick for a first build.