Brett Adams

[Brett Adams @Bre77au]

@kunchenguid I tried it too, and yep, six valid edge case bugs identified and fixed

[Brett Adams @Bre77au]

@bcherny Guess I'm rewriting my entire stack in Go with Fable 5.2 tomorrow.

[Brett Adams @Bre77au]

I accidentally made a huge discovery in June, and it flew under the radar for months, but we know how to control the charging and discharging of Tesla Powerwalls now, both locally and through the cloud.

[Teslemetry @Teslemetry_com]

Teslemetry is one of the first consumer platforms in the world can directly control the charging and discharging Tesla Powerwalls. Take back control and get automating with Teslemetry. https://t.co/Y9EhXpwH1F

[Brett Adams @Bre77au]

How Splunk really licenses metrics: I tested every release since 8.0 https://t.co/yiBFZiQWlC via @LinkedIn

[Brett Adams @Bre77au]

@theo Im not the best dev, but I cant move to T3 yet because I still use agents for more than writing the code, my agents also monitor, patch, and orchestrate deployments.

[Brett Adams @Bre77au]

@MichaelBSeaman @omarvvvr Which technically Stripe has now but when I compared it to others it was much more expensive.

[Brett Adams @Bre77au]

@omarvvvr Do you want to handle your own international tax or just turn a blind eye? Stripe.

Otherwise Lemon Squeezy or Paddle.

[Brett Adams @Bre77au]

@kunchenguid Anthropic hit the same limit and built their own, and it seems many in industry are working on alternative.

[Brett Adams @Bre77au]

@theo I use sometimes use a terminal to prompt my agent, but its the one that prompts my agents.

[Brett Adams @Bre77au]

@fadereeno @omarvvvr @t3dotcodes @theo T3Code looks like another Conductor type worktree organiser, I prefer having full agentic orchestration with one agent controlling all others.

[Brett Adams @Bre77au]

@Frenck And it will make Home Assistant the first consumer platform capable of directly charging and discharging a Tesla Powerwall at will.

[Brett Adams @Bre77au]

@milancermak @kunchenguid I do, it does. Every week my Firstmate reviews browser replays/traces and identities issues real users had.

[Brett Adams @Bre77au]

@LDT0545 Ive never liked the code Codex produced. Over engineered and doesnt follow repository norms.

[Brett Adams @Bre77au]

@ElitzaVasileva I run a complex system comprising of multiple repos. A 20x sub is about the right size for me, I usually have multiple new features built simultaneously, have an agent watching alerts (SRE), and another doing deployments, with one orchestrating all the others.

[Brett Adams @Bre77au]

@TheLLMWhisperer @kunchenguid @poteto What did you move to? I've tried moving away from Firstmate but nothing is as capable at dealing with so many different things and handling events in real time. I look forward to a version of Firstmate that isnt chat based.

[Brett Adams @Bre77au]

I burnt 14.4 billion input tokens, 36 million output tokens, in 169 agent hours over 5140 sessions to ship 466 pull request and launch https://t.co/xfMpdHzmWd in the last 7 days.

Used my Claude Code reset to doubt quota and had @myfirstmate (by @kunchenguid) orchestrating.

[Brett Adams @Bre77au]

@kunchenguid It depends which one is better, and which one is cheaper, because its tradeoffs all the way down.

[Brett Adams @Bre77au]

@kunchenguid Thanks for posting this before my @coderabbitai trial ended, it looks neat and solves the biggest problem with my Firstmate being more productive than I am.

[Brett Adams @Bre77au]

I cant wait to use Claude Haiku 5.5 to spell check my Claude Opus 5.5 prompts controlling Claude Sonnet 5.5 agents.