• for you
  • latest
  • Reviews
    • READERS’ CHOICE
    • ALL REVIEWS
    • ━
    • SMARTPHONES
    • CARS
    • HEADPHONES
    • ACCESSORIES
    • LAPTOPS
    • TABLETS
    • WEARABLES
    • SPEAKERS
    • APPS
  • screen time
    • tv & movie reviews
    • SPOTLIGHT
  • GAMING
    • GAMING NEWS
    • GAME REVIEWS
Reading: Google’s new AI agent never stops fixing your app after launch
Font ResizerAa
Font ResizerAa
  • for you
  • latest
  • picks
  • trends
  • stories
    • TECH
    • AUTOMOTIVE
    • GUIDES
    • OPINIONS
  • REVIEWS
    • READERS’ CHOICE
    • ALL REVIEWS
    • ━
    • SMARTPHONES
    • CARS
    • HEADPHONES
    • ACCESSORIES
    • LAPTOPS
    • TABLETS
    • WEARABLES
    • SPEAKERS
    • APPS
  • screen time
    • tv & movie reviews
    • SPOTLIGHT
  • GAMING
    • GAMING NEWS
    • GAME REVIEWS
  • quick bytes
Follow US

Google’s new AI agent never stops fixing your app after launch

MAYA A.
MAYA A.
36 seconds ago

Google wants its AI agents to keep working on software after it ships. Most coding agents finish their job once a pull request is merged. Stitch Loop, a new experiment from Google Labs, is built to stay connected to a live product, watch how it performs and keep suggesting fixes until a set goal is reached.

Google describes Stitch Loop as a long-running product agent. Teams connect it to GitHub and to their product data, such as logs, user behaviour and customer feedback. They can do that through more than 180 integrations, custom MCP servers or local CLI uploads. The agent then runs ongoing “loops,” each aimed at improving a measurable outcome. Three loops come built in, covering security, latency and accessibility. Teams can add their own, such as one for improving checkout conversion.

Each loop works the same way. Stitch Loop points out an opportunity and links it to evidence from the product. It then proposes one or more fixes, with tradeoffs, scope and design options where relevant. Once a developer chooses a direction, the agent writes the code in an isolated cloud sandbox and opens a GitHub pull request, sometimes with screenshots or video of the change. It then tracks live metrics to see whether the change helped, and it keeps monitoring after the goal is met in case performance slips again. Nothing merges without a human approving the pull request, and Google is careful to stress that point.

The security loop is the most concrete part of the announcement. It combines Gemini 4 Carbon’s security capabilities with CodeMender, the autonomous code-security agent from Google DeepMind, to scan a codebase for vulnerabilities and propose fixes with supporting evidence. Developers who already get automated dependency alerts and patch pull requests will recognise the idea. The difference is the scope: Google says the agent works across the whole codebase, not just packages with known problems.

Google is also positioning Stitch Loop as a layer other tools can use. Coding agents, including Google’s own Antigravity, can pull its product context through the Stitch CLI and MCP, and teams can get daily digests or start new loops by tagging @stitch in Slack.

Several questions remain open. Google hasn’t said what the service will cost. It also hasn’t explained how the agent separates the effect of its own changes from normal swings in traffic and user behaviour, which is much harder to do than fixing code. Companies will also have to decide whether they’re comfortable giving one tool access to their production metrics, user feedback and source code.

For now, Stitch Loop is available only through a waitlist that opens globally today, with priority access for eligible Google AI Ultra subscribers. Lean teams will be the ones to show whether a constantly running product agent saves them time or just adds more pull requests to review.

Follow US
Caffeine. Chaos. Content.
© Absolute Geeks Media FZE LLC 2014–2026.
Proudly made in Dubai, UAE ❤️
Contact · About · Editorial Policy · Privacy Policy
Welcome Back!

Sign in to your account

Username or Email Address
Password

Lost your password?