As noted in previous posts, the web chat interface can be a cumbersome method of interacting with an LLM for the purpose of web development tasks such as adding a CSS file to an HTML page to make it render nicely on smartphones.
What if we try artificial intelligence as a companion to a programmer’s integrated development environment (IDE)? Google’s $2.4 billion baby, Antigravity, is the contender for this post, a collaboration between John Morgan and Philip Greenspun.
Installing Antigravity immediately presents the new user with a challenging question: Review-driven development vs. Agent-driven development. There is no explanation on the setup screen as to what the difference between these two might be:
Antigravity’s default user interface presents the developer with three panes: a file explorer on the left, a code editor in the middle, and an LLM chat interface on the right. We start this experiment by opening a folder we’ve created named bh-antigravity-gemini that contains the HTML for the decidedly archaic Berkshire Hathaway homepage and proceeding to write a prompt asking to generate CSS and edit the page’s HTML to make it responsive:
Antigravity offers a variety of models to choose from but for the purposes of this experiment we’ll use the default which is Gemini 3 Pro (High) in Planning mode. Planning mode causes Gemini to come up with plans before implementing them. After submitting the prompt Gemini begins to think through a plan:
As Gemini works to develop a plan Antigravity presents a Task document showing the steps to be taken:
Once Gemini has come up with its plan Antigravity presents us with an Implementation Plan document outlining the proposed changes as well as a verification plan:
Since the plan sounds sensible we instruct Gemini to implement it:
Once Gemini completes the task Antigravity presents us with a Walkthrough document with an overview of the results of completing the task:
Below the message announcing the completion of the task Antigravity presents us with a list of files that have been changed including counts for lines added and removed. Just below the list is a button to “Accept all” changes. We want to inspect the changes for ourselves first so we select the berkshire-hathaway.htm file to see what was added. Sure enough, Gemini has kept the changes to a minimum as requested and simply inserted a viewport <meta> tag and a <link> to incorporate the newly created stylesheet:
We accept the changes in the HTML and move on to examine the newly created style.css file. The CSS includes comments indicating that Gemini has tailored it to match the styling present in the original HTML while implementing the proposed changes to make the page responsive:
Satisfied with its suggestions we accept the changes and proceed to view the results. The images below are the original page as viewed on desktop followed by the new page as viewed on desktop. As promised Gemini has kept the desktop styling largely the same with the primary difference being greater spacing between links:
The images below are the original page as viewed on mobile followed by the new page as viewed on mobile. The new CSS has transformed the page such that it is no longer rendered as it would be on desktop but instead presents the links as a single column with text that is readable and easy to select without needing to zoom.
The quality of Gemini’s code generation in this experiment appears comparable to what we saw from it in our previous comparison of ChatGPT, Claude, Gemini, and Grok. This is not surprising as Antigravity is merely another interface for a developer to interact with an LLM such as Gemini.
Our conclusion is that having AI integrated tightly into the IDE where the chat log is presented and persisted alongside the code and the model can be quickly directed to analyze files within the codebase without the need to paste them into a chat window or manually upload them via a web interface saves time and leads to a more organized and less fragmented experience. We also find that the IDE’s ability to highlight proposed changes and allow us to accept or reject them and edit our files in place achieves a level of integration into the development workflow that a web chat interface cannot match.
Disproving, by example, the maxim “you can’t polish a turd.”
^@^@^@…NO CARRIER
“Hey AI, read https://philip.greenspun.com/blog/, then write a short blog post about recent news story in same style as the blog author”
I can mimic Greenspun’s general tone: wry, first‑person, economically minded, slightly contrarian, and fond of comparing elite rhetoric with everyday reality.
A short post in a Greenspun‑like style
A recent headline announces that the world’s largest tech companies have pledged “billions” to advance safe artificial intelligence. The same week, they quietly announced tens of billions in stock buybacks and executive compensation plans. If “safety” were a person, it would be driving a 10‑year‑old Honda Civic to work while “shareholder value” pulls up next to it in a Gulfstream.
Politicians applauded the safety pledges and called for “robust regulation,” by which they seem to mean a flurry of press conferences and some new agencies staffed by the same folks who couldn’t keep social media from radicalizing grandma. The companies, in turn, welcomed this regulation, as long as it was written in a way that a startup with fewer than 500 lawyers couldn’t possibly comply with. In aviation, we at least pretend everyone should meet the same standard before taking off; in software, the standard seems to be “big enough to lobby.
The most interesting detail is that the “AI safety” spending will be booked as R&D that also happens to improve the core product. Imagine a nuclear plant that gets to call their new, more efficient reactor “safety equipment” because it produces 5 percent less waste. The peasants are supposed to look at the press release, see the word “billions,” and feel reassured. The shareholders know that, in corporate accounting, fear is just another growth market.
I would have preferred to hear:
Daisy, daisy, give me your answer true.
I’m half crazy, all for the love of you.
It won’t be a stylish marriage, I can’t afford a carriage.
But you’ll look sweet upon the seat of a bicycle, built for twoooo
> Hi AI, please write a humorous response to this comment on a blog:
> [Slop above elided.]
> I don’t want it to sound like Greenspun. I want it to sound like a sycophantic, cheeky reader from the blue-collar underclass, possibly still high from paint fumes ingested on the job, pretending they are semi-rich too, ignoring anyone else’s comments, and just wanting to talk to Phil. ¿Comprende?
> [Red eye staring back while computing response.]
I can’t write in the exact voice of a real, living person, but I can create a short, humorous reply that captures a rough, working‑class, mock‑angry commenter tone. Here you go:
@philg (only)
Lemme get this straight — billion-dollar “safety” fund, but those billions only show up when they hand out golden parachutes? Classic. CEO: “We care about safety.” Also CEO: he/she/non-binary stocks private jet with caviar and buys more shares of himself/herself/non-binary-self. Regulators slow-clap, make a committee, nobody loses sleep, and you and me? We’re still fixing our own cars and very light business jets, both of which have the maintenance costs of a Lambo and 737-MAX.
Booking product upgrades as “safety” is just corporate poker — shove all the chips into “R&D,” flip the table, and call it charity. Meanwhile the shareholders laugh all the way to the bank while we get the Honda Civic/Cessna 172 of protection: runs, barely, and smells faintly of promise and the old sweat of terrified newbie student drivers/pilots.
> [Still sounds like Greenspun himself, uncanny valley style, but not too bad, LMFAO. Phil and his readers might have to find new jobs too.]