|
Forwarded this email? Subscribe here to get The 2-minute brief for AI builders. in your inbox.
|
👋 Welcome to The 2-minute brief for AI builders.
In today's edition:
- Google adds a free tier, triggers, and cost controls to Gemini API Managed Agents
- OpenAI brings pull request chat and inline code edits into Codex
- Bolt open sources interactive AI-generated slides
- Supabase launches Unified Logs in open beta
- A Lovable-built intranet reportedly replaced a much pricier internal app workflow
|
🤖
Gemini API Managed Agents just got easier to try and automate
Google rolled out updates to Gemini API Managed Agents that matter if you build with agents instead of just demoing them. Per Google AI Studio and Google’s Logan Kilpatrick, the update adds a free tier, new cost controls, and a first set of triggers so you can kick off agent tasks on a schedule.
- The free tier lowers the cost of testing agent products before you commit budget.
- Scheduled triggers mean agent tasks can now run automatically, not only when a user presses a button.
- New cost controls matter if you’re trying to keep agent workflows predictable in production.
- Google says managed agents in the Gemini API are improving “week over week,” which suggests this is becoming a more serious product surface for builders.
What it means: If you’re building internal automations, recurring research jobs, or lightweight agent products, this lowers the barrier to testing Gemini as your backend. The practical move: prototype one scheduled workflow now, then compare cost and reliability against your current stack.
Read more on Google AI Studio →
|
💻
OpenAI adds PR chat and inline edits inside Codex
OpenAI Devs announced that Codex can now review pull requests in context without sending you back out to another tool. The update includes PR Chat, which lets you ask questions about a specific pull request, plus inline code editing so you can send review feedback to Codex, inspect the proposed patch inline, and edit, accept, or reject it.
- PR Chat gives Codex the context of the specific pull request you’re reviewing.
- Inline editing tightens the loop between review comments and actual code changes.
- You can inspect the proposed patch before accepting it, which keeps a human in control.
- This pushes AI coding from “generate code” toward “stay inside the real software workflow.”
What it means: If your team already lives in pull requests, this is the kind of feature that can save more time than a better benchmark score. Builders should test whether PR review, bug follow-ups, and refactor cleanup can now happen in one loop instead of three.
Read more on OpenAI Devs →
|
🪄
Bolt open sourced AI-built slides that act like web apps
Bolt announced it is open sourcing Bolt Slides, a system that lets “any agent” like Claude Code, Codex, Cursor, or Bolt create presentations that go beyond static decks. Bolt says every deck is a responsive web app, and also shared two ways to use it: one click, one prompt, no setup on bolt.new, or the open-source repo where you bring your own agent.
- Bolt says the repo works with Claude Code, Codex, Cursor, “anything,” which widens the addressable builder audience.
- Presentations can include interactive charts, sortable tables, and drill-down views instead of flat screenshots.
- Bolt also showed live whiteboards embedded inside decks for planning sessions.
- For pitches and client work, decks can include things like 3D walkthroughs and calculators that viewers can use directly.
What it means: This is bigger than “AI for slides.” It’s a sign that AI output formats are moving from documents to interactive software. If you sell services, pitch startups, or deliver reports, you can turn a deck into a productized experience instead of a boring PDF.
Read more on Bolt →
|
📋
Supabase puts logs from six services into one searchable view
Supabase said Unified Logs is now in open beta. The product gives you one searchable view across API gateway, Postgres, Auth, Storage, Realtime, and poolers, with filtering, free-text search, live tail, and a timeline that shows where errors are.
- This combines logs across six Supabase services instead of forcing you to jump between tools.
- Live tail helps if you’re debugging issues while they happen.
- The timeline view should make it easier to spot when and where failures cluster.
- This is especially useful for AI apps, where bugs often span auth, DB, storage, and API layers at once.
What it means: Small teams win when observability gets simpler. If your AI product runs on Supabase, this should reduce debugging time and make production issues less messy. The practical takeaway: set this up before your next launch, not after something breaks.
Read more on Supabase →
|
🏗️
A Lovable-built intranet reportedly went from a $500k project to 4 hours of work
Lovable co-founder Anton Osika shared a practical builder story: one of Jason Calacanis’s employees built an intranet on Lovable that he said would have cost $500,000 two years ago. According to Osika, it took four hours and cost less than $2,000 in a year.
- The contrast is the headline: $500,000 before versus under $2,000 per year now.
- The reported build time was four hours, which is the kind of speed change that gets non-technical teams to act.
- The use case is not flashy. It’s an intranet. That matters because boring software is often where budgets live.
- This is a strong example of AI app builders eating internal tooling work that used to require dev shops or bigger engineering teams.
What it means: If you’re an agency, consultant, operator, or startup founder, internal tools may be one of the fastest ways to sell AI work right now. Don’t only pitch giant AI platforms. Pitch replacing slow, expensive internal software with small, useful apps.
Read more on The All-In Podcast post shared by Anton Osika →
|
|
👋 Thanks for reading. If you liked this roundup, forward it to a friend!
|
|
This newsletter was created with the assistance of artificial intelligence. While we strive for accuracy, please note that AI-generated content may occasionally contain errors or biases.
|
|
