docs: three-way review vs Claude Code and Codex - #216
Merged
Conversation
A source-verified assessment of where DeepCode stands after 0.2.0, used to set the next round of priorities. Everything in it was checked against the code rather than against BEHAVIOR_PARITY.md, which has drifted before. Headline: the kernel and the 0.2.0 app-server architecture are competitive; the gaps are (a) the CLI presentation layer, (b) sandbox-off-by-default plus a single `mode` conflating permission profile and approval policy, and (c) capabilities that exist in the protocol but have no UI outlet. Also files 15 concrete findings (dead --help flags, wrong --bare copy, version skew, parity wording, CLAUDE.md not read, ...) that the follow-up PRs work through. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Source-verified assessment of DeepCode after 0.2.0, against Claude Code and Codex — capabilities, interface, and where the next round of effort should go. Lands as
docs/THREE_WAY_REVIEW.md+ a README pointer.Why
BEHAVIOR_PARITY.mdhas drifted behind the code more than once, so every row here was checked against the source instead. Baseline verified on this tree:pnpm typecheckpasses,pnpm testexits 0.Headline
The kernel (tools, hooks, skills, plugins, MCP, sessions, sandbox, sub-agents, cron) is competitive with Claude Code, and 0.2.0 added the expensive part — app-server, Thread/Turn/Item, thin clients, credential boundary. The remaining gaps are not feature-list gaps:
y/n/a.modestill expresses both permission profile and approval policy — the split the alignment plan called for in §5.5.workspace/diffand review apply/revert exist in the protocol and in VS Code, but the desktop has zero entry points for them, and zero slash commands (the PlusMenu still says "palette lands in v0.2").Also filed
15 concrete findings with source anchors, including: five
--help-documented flags that are parsed and never consumed;--barecopy describing a different feature; version fields still at 0.1.6 while the changelog says 0.2.0; three "Claude Code parity" strings contradicting the README's current positioning;CLAUDE.mdnot read by the memory loader.Follow-up PRs work through these in the priority order the doc proposes.
Review notes
Doc-only — no code paths touched. The Codex column is grounded in its stable public design and carries an explicit confidence note; none of the conclusions depend on a single command-level detail.
🤖 Generated with Claude Code