ESP32 Panel with Three AIs: Claude Code, Codex, and Antigravity Compared on Habr

Habr author Yadulla Abidi compared three AI code-writing assistants — Claude Code, Codex, and Antigravity — on a real hardware project: building an information panel based on an ESP32 microcontroller with an OLED display and temperature sensor.

What Task Was Set for the Assistants

In 2026, there are many AI code-writing helpers, and they behave differently depending on the type of task. A model capable of whipping up a flawless website in a couple of minutes can get stuck hard on a complex commercial coding project. But for small amateur microcontroller schemes, such models, by the author's observation, are almost ideal — this is where they can show themselves in all their glory.

That is why Abidi chose for comparison a simple but non-trivial task: assemble an information panel on ESP32 with an OLED screen and temperature sensor that, at the same time, should be able to perform several parallel tasks simultaneously — that is, work not as a linear script, but as a full-fledged multitasking firmware.

How Claude Code, Codex, and Antigravity Handled It

Three assistants were pitted against each other: Claude Code from Anthropic, Codex from OpenAI, and Antigravity. All three tools ultimately managed to produce a working prototype of the panel — that is, they coped with the basic task of writing firmware for a microcontroller from scratch.

• Project — information panel on ESP32 with OLED display and temperature sensor
• Requirement — support for multiple parallel tasks in the firmware
• Compared Claude Code (Anthropic), Codex (OpenAI), and Antigravity
• All three assistants coped and produced a working prototype

But, as the author of the review emphasizes, the fact that all three models coped with the task does not mean that choosing an assistant for such projects can be done blindly — the path to a working prototype and the quality of the final code differed noticeably among the assistants.

What This Means

The comparison shows that AI code-writing assistants in 2026 are already mature enough to cover niche hardware tasks — assembling firmware for microcontrollers, which previously required specialized embedded developers. But this maturity is uneven: for amateur electronics projects, the tool still needs to be chosen consciously, rather than relying on whatever comes first to hand.