This Week's AI Tools: Meridian, Omni, Vendo, Culpa

Four founder-facing AI launches from mid-August 2026, plus a model note.

Meridian — on-device AI work journal (open source, free). Writes your day in plain English and drafts standup updates for tools like Jira; you approve before anything posts. No cloud, no account. Worth a try if privacy and daily updates both matter. 1 · meridiona.com

Omni by xpander — move laptop-bound agent workflows into scheduled, shareable cloud agents. Describes wiring tools, mock testing, and keeping agents healthy after deploy. Worth a try if your best loop dies when the lid closes. 2 · chat.xpander.ai

Vendo — open-source embedded layer so customers build views and micro-apps on your API, inside your guardrails and brand. Built for SaaS teams drowning in customization requests. Guardrails still decide safety. 3 · vendo.run

Culpa — local-first AI cost forensics. Traces spend to the user, feature, conversation, and retry loop; forecasts feature cost before you ship. Worth a try when the AI bill is a real line item. 4 · getculpa.com

Gemini 3.7 Flash — Google’s workhorse model update for coding and agents (Aug 13). Intro pricing through end of 2026: $0.75 / 1M input tokens, $3.75 / 1M output tokens. Re-benchmark on your own tasks before trusting vendor charts. 5