Codeium: what owners actually say

Multiple users found the Terms of Service language around 'Autocomplete User Content' alarming, interpreting it to mean code written while using the tool becomes Codeium's property. One user called it 'the catch' for a free product.

Users across HN and Lemmy express distrust about what happens to their code context when sent to Codeium's servers. One user sandboxed the binary and found it downloads a large model and starts a local web server with firewall activity on install. Others note company policies prevent adoption because code is sent externally.

Repeated skepticism about how a compute-heavy ML service can be 'free forever.' Users note the pricing page only says 'early users get free forever access,' leaving later adopters uncertain. The company cites prior ML infra experience but users remain unconvinced.

A user who switched from Copilot reports Codeium doesn't read project-wide context as intelligently — e.g., not picking up how promises are handled across a project the way Copilot does.

Users working in Elixir and Crystal report weak completions compared to mainstream languages, and request broader language coverage. Copilot is described as similarly limited but the gap is noted.

Multiple users report completion quality is on par with GitHub Copilot, with one calling it 'about equal' and another saying suggestions 'just work' after switching permanently.

Owners appreciate availability across VS Code, Vim/Neovim, PyCharm, and Android Studio — wider than Copilot's ecosystem at the time of these comments.

Users highlight completion speed and note visible quality improvements over weeks of use as the team iterates.

Owners praise the developers for quick responses on Discord, frequent updates, and direct engagement on HN and Product Hunt threads.

Several users specifically value Codeium as an alternative to a Microsoft-owned product, citing independence from big tech as a reason to switch.