A lot has been written about how agentic frameworks like Antigravity can boost coding productivity. However, Antigravity isn’t only beneficial for people who spend their days coding. Antigravity can make a real difference for PMs. I’m a PM at Google. Every day I need to handle unexpected challenges, some of which are super-urgent. More and more, I have been tapping into the capabilities of Antigravity to get high-quality solutions for these challenges. In this article I’ll go over a couple of challenges I’ve faced recently, how I would have tackled (or been tackled by) these challenges pre-Antigravity, and how Antigravity helped me to keep moving.
Removing outdated videos from documentation topics
The first challenge I’d like to share involves finding and removing out-of-date videos from technical documentation topics. The documentation for Google Cloud at https://docs.cloud.google.com/docs includes topics that embed videos, like this one: https://docs.cloud.google.com/bigquery/docs/geospatial-visualize-colab. These videos help to illustrate key points from the documentation topics and give a dynamic introduction to concepts that the doc topics elaborate on. In all, videos enhance your experience with the documentation.
What happens when the content of the videos and the text in the documentation topics drift apart? Some older videos are still relevant, but when they include out-of-date product details, they can undermine the overall experience. Earlier this year, this challenge landed on my virtual desk. What could we do to identify and deal with documentation topics that included embedded videos that were no longer current?
The scope of the problem is big. Over 500 Google Cloud documentation topics embed videos. In the pre-AI era, an intern could have easily spent a long (and not entirely pleasant) term going through these topics, manually checking each video, and coming up with an action plan to deal with outdated videos. Let’s say we tried to tackle this problem a couple years ago, prior to agentic frameworks like Antigravity. Back then, it would have taken a solid couple of weeks to come up with a custom harness (like a Jupyter notebook) to prompt a model to detect problem videos and then do the manual work to create PRs to fix the problems that the harness detected.
Lucky for me, in the agentic AI era, I can use Antigravity to help solve this problem. First, I prompted Antigravity to detect all the documentation topics that embedded videos:
Find all the markdown files in <source path> that embed videos.
Save the results in a markdown table here: <link to my MD file>.
In the table have three columns:
Source - links to the source files for all the topics that embed videos
external URL net - links to the external topics on
https://docs.cloud.google.com/ for all the topics that embed videos
video link - direct link to the embedded video
Once Antigravity had detected all the topics containing embedded videos, I prompted it to find the subset of these topics that contained a problematic video:
Now, for all the videos in <link to my MD file>, identify all the videos
where one of the hosts no longer works at Google.
Add a column to <link to my MD file> called “Xoogler” with “YES”
if one of the hosts no longer works at Google and “NO” otherwise.
Of the original 500+ videos, Antigravity identified about 60 that could be out-of-date. Now I could examine a spreadsheet version of the markdown file to see what topics and videos had been identified as out-of-date and do a spot-check on some of the videos.
Once I was satisfied with the videos that Antigravity had identified, I prompted it to open issues (called bugs in Google) to get the problematic videos removed:
The results of the run to find problem videos have been summarized and
vetted here: <link to my MD file>.
For the entries in this file with “external URL net” = “https://docs.cloud.google.com/bigtable/docs/media”,
create a bug with the following characteristics:
Component: %placeholder component%
Title: Video Cleanup: remove expired videos from topic %topic_name%
Where %topic name% = the title of the topic linked to in “external URL net”
Priority: P3
Type: bug
Description: This bug is opened automatically to clean up Cloud doc topics
that embed expired videos (videos that have a Xoogler host or that were
created for a very old event, like Next 2018).
Please remove the link to this video, and any surrounding text that refers to the video, from the topic. Thanks!
Topic: %external URL net%
Videos: %list all values of “video link” for this topic%
Actual component: set value to the component id of the topic
Finally, once I had spot-checked a set of the bugs to ensure that they were formatted correctly and would make sense to their recipients, I asked Antigravity to move each bug from the placeholder component to its Actual component, which it had derived in the previous step.
Following these steps, and allowing for some missteps on my part, I completed the entire task in a couple of hours, including the time spent by Antigravity responding to prompts. Compare that result to 4 months, a decade ago, or two weeks a couple of years ago. Antigravity is, indeed, looking like a PM’s best friend, but before we draw any conclusions, let’s look at how Antigravity helped with another, very urgent, problem.
Using Antigravity to turn around an urgent request
One of my responsibilities is to oversee the delivery of the agent skills in this repo: https://github.com/google/skills. One Friday afternoon in July I got an urgent request — for each of the skills in the repo, answer the following questions:
- Does the product associated with the skill have an MCP server?
- Are any MCP servers explicitly referenced in the skill?
This information exists. You can find the list of products with MCP servers here https://docs.cloud.google.com/mcp/supported-products#google-cloud-mcp-servers and that list can be correlated with the products associated with the skills. For each skill, you can look at the SKILL.md to see if an MCP server is used. So, it’s an easy problem to solve. All I needed was:
- A week to manually go through each of the 80 or so skills to check for MCP references and correlate the MCP server list with the products associated with the skill, OR
- A time machine to go back to the beginning of the project so that we collected this metadata for each skill when it was created.
Since I had neither a week nor a time machine, I decided to turn to Antigravity. Starting with a spreadsheet that contained links to each of the skill definitions in https://github.com/google/skills, I asked Antigravity to create a column to list the MCP servers used by each skill:
update <link to my spreadsheet> so that the "mcp servers used" column
lists the MCP servers used by the skill in that row
Next, I asked Antigravity to update the spreadsheet to add a column to show whether an MCP server existed for the product associated with each skill:
now update the spreadsheet again.
Only change the "MCP server exists for product?" column and change nothing
else in the spreadsheet.
If an MCP server exists for the product according to https://docs.cloud.google.com/mcp/supported-products,
put YES in the "MCP server exists for product?" column.
Otherwise, put NO in that column.
With that, I had an answer to the urgent, Friday afternoon question. Instead of taking me a week, it took me about 20 minutes, allowing for a couple of iterations to get exactly what I needed. I shared the spreadsheet that Antigravity updated and I was done — Friday afternoon crisis averted!
What I learned from using Antigravity to solve PM problems
I’ve shared just two of the ways that I streamline my PM work, and get things done faster, using Antigravity. Here’s what I’ve learned so far:
- As soon as I catch myself doing a repetitive task (for example, opening a set of child bugs), I stop myself and ask, “could I use Antigravity to automate this task?”
- When I need to correlate two or more data sources (like the skills repo and https://docs.cloud.google.com/mcp/supported-products#google-cloud-mcp-servers in the second example), I stop and ask myself, “could I use Antigravity to correlate these data sources?”
- When I need to use a tool that has a challenging interface or that I have forgotten how to use, I stop and ask myself, “could I use Antigravity as my go-between with this tool?”
The most-hyped use cases for agentic frameworks like Antigravity may be code generation, and I’ve certainly benefited from Antigravity’s code gen capabilities, but in my day-to-day work as a PM, I am constantly finding new ways to harness Antigravity to tackle PM problems that would have either been prohibitively time-consuming or downright impossible. I can honestly say that for me, as a busy PM, Antigravity is my new best friend.
Antigravity: the busy PM’s best friend was originally published in Google Cloud – Community on Medium, where people are continuing the conversation by highlighting and responding to this story.
Source Credit: https://medium.com/google-cloud/antigravity-the-busy-pms-best-friend-7bae6e8a00bf?source=rss—-e52cf94d98af—4
About the Author
