bazhand · HackerTrans

11 gün öncediscuss

It was something that was used for token efficiency. Most of the settings and use cases are quite poorly communicated but asking Claude to review the latest release changelog (https://github.com/anthropics/claude-code/blob/main/CHANGELO...) is quite useful. Combined with @"claude-code-guide (agent)" to read it's own docs for settings/configs is super helpful.

The quite useful tool is to use /opusplan along with /codex:rescue (https://github.com/openai/codex-plugin-cc) means you get quite a strongly reviewed plan using native claude + codex without having to implement the mostly useless trust-me-bro plugins and other bs.

bazhand

11 gün öncediscuss

Have you tried '/model opusplan' I've had strong results mixing opus for planning with sonnet implementing.

bazhand

12 gün öncediscuss

Been using https://cmux.com for past few months, and runs nicely over ssh + tmux with native tabs.

bazhand

2 ay öncediscuss

/advisor is like the old /opusplan mode but for running tasks not just pre-planning. It can work nicely with Sonnet as the main agent and escalates to Opus as needed.

bazhand

3 ay öncediscuss

export CLAUDE_CODE_EFFORT_LEVEL=max

bazhand

4 ay öncediscuss

and now /fast will get you moving quicker but at a premium.

bazhand

4 ay öncediscuss

I ran into this exact problem building a MCP server. 85 tools in experimental mode, ~17k tokens just for the tool manifest before any work starts.

The fix I (well Codex actually) landed on was toolset tiers (minimal/authoring/experimental) controlled by env var, plus phase-gating, now tools are registered but ~80% are "not connected" until you call _connect. The effective listed surface stays pretty small.

Lazy loading basically, not a new concept for people here.

bazhand

4 ay öncediscuss

this is the vibe coding equivalent of 'hello world' at this stage. Fun tho.

bazhand

5 ay öncediscuss

The third most popular skill on skills.sh[1] with 50k/week installs is a link to download a command[2]

[1] https://skills.sh/vercel-labs/agent-skills/web-design-guidel... [2] https://github.com/vercel-labs/agent-skills/blob/main/skills...

All of these SKILLS.md/AGENTS.md/COMMANDS.md are just simple prompts, maybe even some with context links.

And quite dangerous.

bazhand

6 ay öncediscuss

FWIW this isn’t new, using a Claude/Max subscription auth token as a general-purpose “API key” has been known (and blocked) for ages. OpenCode basically had to impersonate the official Claude Code client to make that work, and it always felt like a loophole that would get patched eventually.

This is exactly why (when OpenCode and Charm/Crush started diverging) Charm chose not to support “use your Claude subscription” auth and went in a different direction (BYOK / multi-provider / etc). They didn’t want to build a product on top of a fragile, unofficial auth path.

And I think there’s a privacy/policy reason tightening this now too: the recent Claude Code update (2.1-ish) pops a “Help improve Claude” prompt in the terminal. If you turn that ON, retention jumps from 30 days to up to 5 years for new/resumed chats/coding sessions (and data can be used for model improvement). If you keep it OFF, you stay on the default 30-day retention. You can also delete data anytime in settings. That consent + retention toggle is hard to enforce cleanly if you’re not in an official client flow, so it makes sense they’re drawing a harder line.

bazhand

7 ay öncediscuss

How does the memory scale (or not!) over time. If using Letta in a single agent mode, similar to just using claude - how does memory blocks stay relevant and contextual?

I guess what I'm asking is, if there is a memory block limit, is that an issue for self learning over time. Claude as you know just straight up ignores CLAUDE.md and doesnt self-improve it.

bazhand

9 ay öncediscuss

I like this approach, and it's very similar to what I'm currently using - albeit my use case is to get Claude to automatically invoke the relevant subagents during both planning and build stages.

Subagents using MCP would solve quite a lot of the context preparation required for a successful session. I'm using GLM-4.6 vision/web search MCP in claude code to relative success as the subagents can access these via API (not true MCP behaviour but close enough)

bazhand

10 ay öncediscuss

Installing (Arch btw), love the UI

bazhand

10 ay öncediscuss

Can it work without Linear, using md files?

comments