You're sitting right there, ready to work, and the power or internet goes down ten minutes before an important call.
First, a key stakeholder is staring at an empty video-call room, concluding that you flaked. Then, hours later, you weed through all the "Hello? Are you there?" emails, and have to send the "so sorry, my internet went out" message, which may be true, but still sounds exactly like "the dog ate my homework" and feels very lame after the fact.
I built proxigenta (proxigenta.com) to relieve both issues. It's a small menu bar agent for Mac, Windows, and Linux. When your machine stops checking in to the remote server, proxigenta notices and emails the people you chose (your manager, your main client, a teammate) with the message you wrote in advance: "proxigenta has identified an outage, and Alex is currently offline."
I'm a developer experimenting with a voice-controlled AI assistant that can operate your computer. The idea is basically: imagine you had an extremely tech-savvy assistant sitting in front of your computer 24/7 that you could call at any time. I use it for programming, but I'm curious if there are other use cases for non-developers. For the sake of this poll, please assume you can trust the AI to make fewer mistakes than a competent human assistant.
A while back I posted here about counting the sources behind a handful of AI answers. I kept going, and I now have 241 software categories measured.
For each category I asked ChatGPT and Gemini the six questions a buyer actually types (best X, best free X, most affordable X, best for small teams, what should I use, recommendations), gave each one a live web search, and logged every product named and every source behind the answer. That is 1,446 answers and 5,037 product mentions.
Three things stood out:
1. Reddit was a source behind the answers in 93% of categories. Capterra, the review directory a lot of us pay for, in 12%. The assistants lean on forums and independent roundups, not the review grid.
I'm excited to introduce FitBuddy an all-in-one fitness tracker I built for people who want to focus on training, not juggling multiple apps.
Like many lifters, I was tired of switching between different apps for workouts, calories, body weight, sleep, and water tracking. I wanted one clean app that made progressive overload effortless while keeping all my health data in one place.
I kept forgetting DSA patterns between interview cycles, but opening a laptop for a full coding session was too much friction.
I built AlgoBite as a smaller review loop: reconstruct a real Python solution from scrambled chunks, run browser tests, then let spaced repetition bring it back before you forget it. It is designed for a five-minute phone session, not to replace solving from scratch.
I built ContractAI because generating a contract is only one small part of actually dealing with contracts.
ContractAI brings the workflow into one place:
AI contract generation AI risk analysis with severity levels and suggested fixes AI negotiation with counter-proposals from your perspective Rich contract editor E-signatures with an audit trail 185+ contract templates Bulk contract generation Team collaboration REST API Stripe billing and template marketplace 5 AI providers: OpenAI, Anthropic, Gemini, Groq and xAI 8 languages
I also made the demo available without requiring an account, so you can see how it works before signing up.
One week they manually search for prospects. The next week they are busy delivering client work. Outbound stops, follow-ups get forgotten, and revenue becomes unpredictable.
Quick follow-up to my intro post here. I'm Ryan, still building Retic solo, and I finally have something better than a screenshot to show.
The whole pitch for Retic has been "check a task, and the goal above it moves on its own." Until now I'd only described that in text. Attached is the actual UI: check the one task left under a goal, and it jumps straight to 100% and gets marked achieved, in the same tree you manage everything else in.
