frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

What I changed in how I use Claude Code after Anthropic's postmortem

4•cinooo•4h ago
After watching Anthropic's recent postmortem (anthropic.com/engineering/april-23-postmortem), I've been thinking about the way I approach Claude Code differently. They lowered default reasoning effort to fix latency, called it the wrong tradeoff, and reverted under public scrutiny. The reverts are good but they don't change the underlying equation I'd been ignoring.

The fact is we potentially have a team of engineers at our disposal now. Token cost is a real cost, but I couldn't hire my own engineer for any of the personal work I'm doing. Apply that lens to token usage and the approach shifts. It stops being about the cost ceiling and becomes a cost/output/quality view, the same way I'd think about hiring decisions on a real team.

Four places I now think about: model, configuration, prompting, agents.

On model. Opus is still the strongest for critical decisions and architectural reasoning. Sonnet is usually good enough for coding and simple repetitive tasks. I use the right model for the work. If I cheap out, I can't expect quality.

On configuration. /effort runs from low to max. Opus 4.7's default is xhigh. I set the level to fit the work. A quick edit doesn't need max, an architectural decision does. The cheapest move and the one I'd been skipping.

On prompting: three patterns I find most effective.

1. "Ask questions if unsure." Without this I'm not giving the model an out, which closes off solutions even when there's no clean answer and tradeoffs need to be surfaced.

2. "Time and cost are not factors here. Prefer robust, sustainable, scalable solutions, do not leave tech debt." This inverts the implicit optimisation pressure for the duration of the task.

3. "Reflect on this session and encode via claude.md or skills what you learned, so the next iteration doesn't repeat the same mistakes." Worth capturing as a skill and iterating for myself. Without this every session starts from zero, repeating mistakes I've already corrected.

On agents. Without going into detail as this is a whole post in itself, the pattern that works for me is using agents to separate concerns. One agent does spec review against the code (code is source of truth). A separate agent does code review after implementation.

Engineering and product teams have always balanced speed to market with cost and quality. AI is no different. The difference is which levers I choose. Spend the budget on effort deliberately, and the work comes back at the level I want.

Tell HN: VS Code v1.117.0 automatically adds GitHub Copilot as your co author

5•adithyassekhar•1h ago•2 comments

Claude.ai down again?

28•zh_code•5h ago•20 comments

What I changed in how I use Claude Code after Anthropic's postmortem

4•cinooo•4h ago•0 comments

Ask HN: Is anyone still using the ChatGPT browser, Atlas?

3•widenrun•1h ago•1 comments

Ask HN: Anyone feel like they're just opting out of tech these days?

13•moomoo11•8h ago•7 comments

Tell HN: An update from the new Tindie team

77•altairprime•22h ago•48 comments

Tell HN: An app is silently installing itself on my iPhone every day

585•_-x-_•4d ago•187 comments

Increase in Claude Code speed bumps?

3•jesse_dot_id•10h ago•0 comments

Ask HN: Hosted Fossil for small teams – interesting, or wrong call?

7•ragelink•10h ago•3 comments

Ask HN: Is it still worth it to try to get a job in IT

7•morpheos137•15h ago•12 comments

Ask HN: Do you think AI will end up like dotcom?

5•dicksent•6h ago•2 comments

Ask HN: If coding gets faster, where should architecture happen?

5•karlosh•19h ago•4 comments

Ask HN: Anyone else suddenly have Brawls Stars on their iPhone?

4•realberkeaslan•19h ago•2 comments

Ask HN: Is it just me or is Claude Code getting worse?

28•e-nouri•1d ago•18 comments

SYNQ – Give ChatGPT and Claude permanent, local memory

2•Tamatarr•9h ago•0 comments

Tell HN: Claude 4.7 is ignoring stop hooks

108•LatencyKills•5d ago•88 comments

Tell HN: One Medical Is a Nightmare

31•rincebrain•2d ago•6 comments

Ask HN: When might we not have to do laundry or fold clothes or cook

23•samarthv•1d ago•19 comments

Why don't AI coding tools like REST?

16•sshadmand•1d ago•5 comments

Ask HN: Anyone want to collaborate on a local-first AI-based research assistant

24•venkatram-s•3d ago•9 comments

Ask HN: Do you read differently now that anything could be AI generated?

33•dwa3592•4d ago•31 comments

You've reached the end!