Principal Product Manager @Adobe . Tweets are my own.
San Francisco, CA
Joined June 2008
- Tweets7.9K
- Following1.8K
- Followers1.2K
- Likes95
Brian Au retweeted
i recently added this command to the claude-api skill. run it in Claude Code to fix common prompting "anti-patterns" that can hobble frontier models:
/claude-api prompt-audit
patterns include:
1/Verification rituals. Instructions like "double-check your work” or "verify twice before responding” are often taken literally by frontier models and can waste tokens.
2/ Thoroughness and emphasis boosters. "Be maximally thorough," "CRITICAL: YOU MUST ALWAYS…" can lead to verbosity and extra tool calls when working with frontier models.
3/ Mandatory procedures and scratchpad scaffolds. Fixed step processes (e.g., "think step by step in a scratchpad") or reasoning templates are rituals that frontier models don't need. This scaffolding can stack on top of native reasoning and use unnecessary tokens.
4/ Stale examples. Few-shot examples tuned to an older model's failure modes can teach a frontier model to imitate long reasoning chains on requests that don't need them.
5/ Contradictory rules. Frontier models are better at instruction following. Contradictory instructions ("always refund within policy" vs. "never issue refunds without escalation") can be followed more literally by frontier models, resulting in degraded performance.
6/ Dated configuration. Settings written for an older Claude generation (e.g., manual thinking budgets) can be rejected by the Claude Platform with newer models.
these patterns accumulate in prompts over time, and can quietly degrade performance when upgrading to newer models. a common reason is the frontier models are better at instruction following, so these anti-patterns steer them to spend unnecessary tokens.
example: i tested a migration from Opus 4.8 to Opus 5 on an internal customer support benchmark. with Opus 5 (and other frontier models like Fable 5.1), verification rituals ("verify twice") use unnecessary tokens by duplicating work. emphasis boosters ("be maximally thorough") become dozens of unneeded searches.
applying prompt audits can improve performance and reduce cost (as shown in example attached and will be sharing a full write-up soon).
also, the skill is also open source and some of this guidance likely applies generally across frontier models
github.com/anthropics/skills…
xitter.cf/petergyang/status/2094…
Brian Au retweeted
This is so interesting because there's a measurable impact from newer/older balls on rates of aces, holds, unreturned, double faults, 1st serve in %, rally length et al. So if balls are changed more often, in theory things like serve dominance may increase (from existing all time highs btw). Impact will likely be small (double faults also rise with newer balls, 1st serve in % decreases modestly, & hold rate changes are smaller). Probably only detectable for season-aggregated unreturned/ace rates and multi-season aggregates for hold rates. But it's probably easy to underestimate 2nd & 3rd order effects supposedly small tweaks like this can have on the way the game is played. Tennis is already massively short point and serve dominant, and changes like this nudge it slightly further in that direction
Brian Au retweeted
The thing about people talking about building with AI is that they always talk about how they’re building, what tools they use, and how much they use them.
Much less is said about what they actually built, or what impact it had.
The tool becomes the job. And the purpose.
Brian Au retweeted
The obvious path for designers in the AI age is to move closer to code.
But the more valuable path may be upstream: closer to the customer, the business, and the problem.
If everyone can prompt agents to code, the scarce skill becomes knowing why, what, and how to build.
Brian Au retweeted
"There's a good looking guy with white hair looking down on us right now that's happy!" 🥹
Tommy Lloyd pays homage to the late Lute Olson in his postgame interview 🙏
#MarchMadness
Brian Au retweeted
For my friends who are still using UV and might be a little weary about recent compromises to PyPi packages, stick this in your pyproject.toml.
You can let all of those pip users find and report the compromises...
Brian Au retweeted
We're excited to announce duckdb-skills, a DuckDB plugin for Claude Code!
We think the embedded nature of DuckDB makes it a perfect companion for Claude in your local workflows.
The skills supported include:
+ read-file and query – uses DuckDB's CLI to query data locally, unlocking easy access to any file that DuckDB can read.
+ read-memories – a clever idea to store your Claude memories in DuckDB and query them at blazing speed.
These are powered by two additional skills:
+ attach-db – gives Claude a mechanism to manage DuckDB state through a .sql file linked to your project.
+ duckdb-docs – uses a remote DuckDB full-text search database to query the DuckDB docs and answer all of your (and Claude's own) questions.
github.com/duckdb/duckdb-ski…
Brian Au retweeted
yeah it is but everything in moderation. Internally we always talked about main quest and side quests.
Everyone should focus on the main quest, and moderately or not all on side quests.
Both quest lines feel productive but only one of them advances the main mission of the company.
Brian Au retweeted
Introducing the Google Workspace CLI: github.com/googleworkspace/c… - built for humans and agents.
Google Drive, Gmail, Calendar, and every Workspace API. 40+ agent skills included.
Brian Au retweeted
In the next version of Claude Code..
We're introducing two new Skills: /simplify and /batch. I have been using both daily, and am excited to share them with everyone.
Combined, these kills automate much of the work it used to take to (1) shepherd a pull request to production and (2) perform straightforward, parallelizable code migrations.
Brian Au retweeted
Announcing a new Claude Code feature: Remote Control. It's rolling out now to Max users in research preview. Try it with /remote-control
Start local sessions from the terminal, then continue them from your phone. Take a walk, see the sun, walk your dog without losing your flow.
Brian Au retweeted
Introducing: built-in git worktree support for Claude Code
Now, agents can run in parallel without interfering with one other. Each agent gets its own worktree and can work independently.
The Claude Code Desktop app has had built-in support for worktrees for a while, and now we're bringing it to CLI too.
Learn more about worktrees: git-scm.com/docs/git-worktre…
Brian Au retweeted
I'm Boris and I created Claude Code. I wanted to quickly share a few tips for using Claude Code, sourced directly from the Claude Code team. The way the team uses Claude is different than how I use it. Remember: there is no one right way to use Claude Code -- everyones' setup is different. You should experiment to see what works for you!