I've been using Antigravity alongside app development and art projects for about a year. Here's the honest assessment — not a feature comparison, but what it's actually like to use as someone who thinks about aesthetics as much as they think about code.
Why I Started Using Antigravity
What I wanted from a coding tool wasn't just code execution — I wanted something that could think alongside me about what I was trying to build. The distinction matters when the output isn't just functional software but something with an intended aesthetic.
The first thing that surprised me about Antigravity was its plan/fast mode structure. Before making large changes, it drafts what it's going to do. That pause felt natural to me — it matched the rhythm of designing something before building it.
Three Things That Have Genuinely Worked
Plan Mode and Fast Mode Match How I Think
The two-mode structure (Plan for architectural decisions, Fast for targeted edits) maps onto how I naturally work. When adding a new feature, I want to think through the implications before touching code. Plan mode gives me that — it shows what will change and where before anything is committed. Then Fast mode handles the execution.
This isn't just an aesthetic preference. It's reduced the number of times I've had to revert something because the AI went further than I intended.
Context Compounds Over Time
Antigravity gets noticeably better the more context you give it. AGENTS.md and KNOWLEDGE.md entries that explain your project's philosophy — not just its structure — shape the quality of suggestions in ways that surprised me.
Writing "this project uses a minimalist design philosophy, monochrome palette, nothing decorative that doesn't serve a function" changed how Antigravity names components, structures files, and comments code. It stopped feeling like I was directing a tool and started feeling like working with someone who understood what I was going for.
Error Diagnosis That Teaches Rather Than Just Fixes
When something breaks, pasting the error and asking what it means tends to get back an explanation of the situation, not just the fix. Often Antigravity will explain why the error is occurring and offer multiple approaches to resolving it, noting the tradeoffs.
Over a year, this has genuinely improved my understanding of the codebase rather than just keeping it working. That matters when you're the only person maintaining multiple apps.
What Still Falls Short
Honest answer: there are real gaps.
Visual and animation instructions are hard to convey: Text-heavy features and logic land well. But "make this animation feel more organic" or "increase the line height just in this section" — sensory, visual instructions — don't reliably translate through natural language. I still often end up editing CSS values or animation parameters directly.
Context drift in long sessions: After an hour or two, Antigravity starts to lose the earlier constraints I set. The KNOWLEDGE.md approach helps significantly, but there's still overhead in maintaining consistency across long sessions.
Japanese code comments: Even with Japanese set as the project language, code comments sometimes default to English. Explicitly writing "code comments in Japanese" in AGENTS.md is the reliable fix, but it shouldn't require explicit instruction.
Why I Keep Using It
The reason I've stayed with Antigravity after a year is harder to quantify than a feature list. Most AI IDEs feel like tools that execute instructions. Antigravity more often feels like something that's trying to understand what the project is for.
That distinction matters to me as someone who cares about the character of what I build, not just whether it works. Whether that's worth the tradeoffs depends entirely on what you're building and how you think about the work.
For solo developers, indie makers, and people who work across creative and technical domains, Antigravity's strengths align well with what those roles actually need. For large team environments or highly complex backend systems, other options might fit better. That specificity isn't a criticism — it's just an accurate map of where Antigravity earns its place.