Cody

Data is a HackerNews beta-era launch thread (19 comments total, incl. 4 trivial ProductHunt tagline remarks). POSITIVE: One multi-month VS Code user reports 'a ton of value': Cody 'really knows how to summarise code blocks' and 'misses very little', is 'very smart when it comes to filling in gaps in log statements, error messages or code comments', and makes copy-paste 'mostly dead' for common patterns. NEGATIVE: Another user 'mostly gave up' — 'no clear instruction about how to add/remove repos, or how to get Cody to recognize the current repo', UI 'poorly planned, with missing, hidden and inconsistent buttons'; he spends more time figuring out Cody than he saves, though he concedes it 'has potential'. NEUTRAL/CURIOUS: One user checked the privacy notice and found it 'generally reasonable' but wants transparency on partner LLM entities (Anthropic?) and downstream terms. Others ask how Cody compares to Copilot (one Copilot power-user calls it near-telepathic, 'hundreds of perfect unit tests') and cursor.sh, whether 'knows your entire codebase' means RAG or full-context loading, and whether training data beats ChatGPT's 2021 cutoff for things like UE5 C++. Several users admit the field is 'gpt variants in skins' with 'no objective basis of comparison'. STAFF REPLIES: Sourcegraph team confirm an experimental Cody CLI, 'bring your own LLM account' for enterprise only (not yet in the editor extension), Claude/GPT-4 under the hood with experimental alternatives, and enthusiasm for Code Llama infill. Editor-extension quality is 'slightly lower' than the desktop app 'but working on parity'. ProductHunt comments are cosmetic: the 'gems' tagline (unrelated to Ruby gems) was renamed to 'nuggets'.

No aggregate ratings were found for this product during the last harvest.