AI agents are showing up in the tools you already use, here's what that means

Table of contents

Technology keeps moving fast, and a lot of it is genuinely exciting. This is where we slow down, pull out what actually matters, and share our own take, and if any of it is something you've run into yourself, we're always happy to talk it through.

This edition: an AI agent that can merge your pull requests, one that organizes your photo library, and one that fixes a decades-old problem in dubbed video, three small updates that caught our attention this time around.

VS Code adds Agent Merge, letting an AI agent fix and merge your pull requests

What happened

Microsoft's September 2 release, VS Code 1.136, introduces a preview feature called Agent Merge. It has an AI agent address review feedback and fix failed checks on its own, continuing until a pull request is mergeable. The release also adds experimental multi-root workspace support for Copilot and Claude agents, along with new enterprise controls around dictation data.

Our take

We flagged this exact shift when Claude Code turned on auto mode by default a few weeks ago: an AI agent's work moving straight through to a merged, shipped result with less friction along the way. Agent Merge pushes that a step further down the pipeline. It's not just generating code anymore, it's resolving review feedback and completing the merge, which means a mistake here isn't caught in a draft, it's already in your codebase.

That's not a reason to avoid the feature. It's a reason to be deliberate about where it applies. The same discipline we wrote about in AI can write code, but it can't guarantee stability holds here: decide which repos, branches, or PR types are actually eligible for an agent-completed merge, and keep an explicit human checkpoint on anything that touches production-critical paths. The convenience is real. So is the need to decide on purpose where you want it. Happy to help you think through where that line should sit.

Source: InfoWorld, "Visual Studio Code 1.136 introduces agent merges for pull requests"

Gemini Spark can now manage and edit your Google Photos library

What happened

Google's personal AI agent, Gemini Spark, can now edit images, curate albums, auto-generate shared albums, and even turn a photographed concert flyer into a calendar event, all inside Google Photos. It's rolling out over the next few weeks to Gemini AI Pro and Ultra subscribers in the US.

Our take

This one's a fun, small example of AI actually saving you a step instead of asking for more of your time. Turning a photographed concert flyer into a calendar event is the kind of detail that's easy to miss until you see it working: a photo you might've just deleted turns into something useful, with the tedious part done for you.

That's the part we find genuinely interesting, not just as an AI story but as a product one. A lot of teams building their own AI features default to a chat box, ask a question, get an answer, when the more useful pattern is often exactly this: AI handling a small, specific task well enough that it just quietly makes things easier. Getting that right is a design problem as much as a model problem. Happy to talk through what that could look like in your own product.

Source: The Verge, "Google Photos is coming to Gemini Spark"

Amazon Prime Video uses AI to fix mismatched lips in dubbed shows

Cozy living room at night with a TV glowing softly and a warm highlight around the screen edge, Amazon Prime Video logo overlay in

What happened

Prime Video launched a feature that combines AI with traditional visual-effects techniques to digitally adjust an actor's mouth movements so they match dubbed audio. It's currently live only on the English dub of the German series Maxton Hall, with more titles planned. Meta and YouTube have rolled out similar lip-sync dubbing for creator content recently, suggesting this is becoming a broader platform trend rather than a one-off experiment.

Our take

This is a good example of AI solving one specific, well-defined problem rather than promising to reinvent an entire product. Mismatched dubbing is a real, persistent, mostly-unsolvable-until-now quality issue, and Prime Video went after exactly that, not a full AI-generated dubbing pipeline, not a broader content overhaul.

That's the pattern we point clients toward when they're figuring out where generative AI actually belongs in their own product. The highest-value AI features usually aren't the flashiest ones, they're the narrow, specific, previously-annoying problems that AI happens to be genuinely good at solving, bolted onto a product that already works. Scoped and specific beats broad and ambitious more often than the AI headlines suggest. Happy to help you find where that scoped opportunity is in your own product.

Source: The Verge, "Amazon Prime Video's AI lip-sync dubbing"

Related: Generative AI services

The pattern across all three

A merged pull request, a smarter photo library, cleaner dubbing. Small updates, but each one makes something people already do a little easier, and that's worth paying attention to.

We like keeping an eye on stories like these as much as the bigger ones. If something here reminded you of a project you're working on, or just made you curious about where AI could fit into what you're building, we'd genuinely love to hear about it.

Let's talk

Share this post: