Yesterday in Launches: AI Agents, Open-Source Notes, and a Laser Uninstaller

If you track new developer tools, yesterday’s batch was a reminder that the category has stretched well beyond code editors and CI pipelines. The five products that launched on September 16 split roughly into two camps: four of them are directly AI-adjacent, and the fifth is a 20-year-old Mac utility that got rebuilt for the 31st century. There’s an AI sales agent, a research agent platform, a privacy-first meeting notetaker, a WordPress plugin for AI visibility, and a delightfully absurd uninstaller. None of them had finalized community vote counts at the time of writing, so consider this a first-look digest rather than a verdict.

Naoma AI Demo Agent V2

Naoma replaces the book-a-demo form with an AI video sales agent that starts a personalized product demo the moment a visitor wants one. No scheduling, no waiting for a rep to be free, no time zone math. The agent walks prospects through the live product, answers their questions, qualifies them, and routes leads into the CRM, a calendar, or checkout depending on intent.

The pitch lands hardest for B2B SaaS teams. The claimed conversion lift, from the typical 1-2% visitor-to-demo rate up to 6-20% with Naoma, is eye-catching even allowing for vendor math. The 33-language support removes a real constraint for global sales teams, and the named customers (Hoteza, AiSDR, UXPressia) give it some credibility.

My honest take: “89% of users say it feels human” is a self-reported stat I’d want to verify myself before betting on it. But the underlying problem is real. A prospect researching at midnight in a language your reps don’t speak is a prospect that never gets a demo today. Naoma at least bridges that gap.

Web Search Agents by Nimble

Nimble’s Web Search Agents are self-learning research agents for specific domains: company enrichment, regulations research, market analysis, and similar repetitive tasks. They crawl the web, accumulate knowledge over time, and return results shaped to a schema you define. The pitch centers on three things: accuracy that compounds past the first query, lower token costs, and auditable Search Plans that show exactly what was searched and why.

This is the most developer-oriented product in the batch, aimed squarely at agent builders. Native integrations with Anthropic, GPT, LangChain, and Vercel reinforce that position.

My reservations: benchmark tables are everywhere in this space and hard to adjudicate independently, and the compounding knowledge claim only pays off if your research task is something you run repeatedly rather than once. The token-cost angle is genuinely compelling, though. Retrieving only what’s needed instead of feeding raw pages through an LLM represents a meaningful saving for teams paying per token.

Oats

Oats is a free, open-source meeting notetaker for macOS that records audio your Mac can already hear. No bot joins the call, no audio leaves your device (unless you opt into the cloud tier), and when the meeting ends you get a summary, per-person action items, and a quality score.

This one is my favorite of the five. The design decisions are all the right ones. Local recording means it works with Zoom, Meet, Teams, phone calls, even hallway conversations. Notes are stored as plain