Show HN: Opus 5.5 made a fully featured Figma clone in 3 hours / 8 prompts

Commenters debated a demo in which Opus 5.5 on Mana produced a Figma-like editor in a few prompts. Some, like xyzsparetimexyz and ConceitedCode, treated it as unsurprising demoability - models can make convincing-looking results but often don't hold up under scrutiny - and asked for concrete prompts, token counts and harness settings (pbk111). An ex‑Figma commenter (jjcm) warned that parity for casual users is believable but power‑user features, variables, and multiplayer/social collaboration contain many subtle gotchas that agents are unlikely to derive correctly. Others voiced skepticism about the marketing angle (anonymous908213) or whether examples were in the training data (kumama), while a few contributors (behzadhaghgoo, behzadhaghgoo again) called it a paradigm shift and credited rapid model improvement and Mana’s platform for enabling one‑shot results.

The conversation split on robustness versus practicality. Martinpw and jjcm argued complex real‑world apps accumulate tens of thousands of lines of battle‑hardened code to handle edge cases, so agents will likely implement common paths first and require long iterative fixes. By contrast bb88 and some Mana proponents said constrained, targeted apps can be vibe‑coded effectively and agents can be used to patch failures over time. Others asked practical follow‑ups (drewbitt, xnx) or requested reproducible prompts and settings, leaving the principal divide between bullish claims of immediate utility and skeptical emphasis on edge cases, collaboration, and long‑term maintainability.