Sheldon ⚡ Hull

@burkeholland Confused on this: - Autopilot seems to consume more credits not just 1 per prompt. - Same as running without permissions? Works best this way but does that mean deny still flags but all other ask/approve is ignored like Claude code with skip permissions?

Underrated feature. It’s a built-in Ralph loop.

@davidfowl I've not been active with dagger lately but feels like that local agent / ci near equality has a lot of value. Simply eliminating environment issues with a centralized preflight that works regardless of OS would be amazing.

CI/CD pipelines are due for a do-over in the agentic era...

@thdxr @mweinbach Skills are great for providing some more semi-deterministic behavior but I'm been saying for months that your approach with SDK tools is what I wish all converged on. Copilot apparently did a similar tool SDK recently. Cut out the extra layer and just give me SDK tools! 🙇

@mweinbach a company who doesn't even have a coding agents was trying to make a standard for plugins based on claude code and get us to adopt it and then cc just added if statements as json fields that we would have had to implement glad we said no, much prefer our code based approach

everyone made a big commotion about skills, a bunch of secondary tooling got built, standards got established and the majority of users have never installed a skill i wish everyone would chill out, most ideas are bad, restraint is more important than ever

it’s a good post. I like to compare ai engineering to ci/cd - everyone wants to optimize it, it burns a lot of time, innovation is good but if everyone is innovating it’s chaos. Can you imagine if every third engineer had their own custom Jenkins server running under their desk?

@burkeholland I've been doing this for a while having a agent stop hook that forces `hk check` or precommit/prek whatever you use. It's a big quality of life improvement. I don't think it's revolutionary, just a reasonable step. Doesn't feel like it solves anything profound though

We cannot continue to let AI write code the way we currently are. It’s way too good at cutting corners. I think hooks in the Copilot CLI might be the answer here. And I think we may only need two of them. preToolUse - the gate: Code gets linted or the model doesn't get to save the file. Model doesn't get to commit until all tests pass. No exceptions. The AI literally cannot move forward without passing basic muster. postToolUse - the feedback loop: Run your type checker after every edit (tsc --noEmit, whatevs). The output goes straight back into the model's context. It sees the errors. It fixes them. Loop restarts and continues until code is actually correct. No rationalizing about why it’s ok for some dumb error to persist. The model cannot write a file, pass a type check, or commit without actually doing what was asked. It’s time we held models to the same basic standards we hold for ourselves.

Fuck it, we just made Fizzy completely free. The open source installable version was always free, but the SaaS version was pay. No more. Basecamp and HEY's largess will subsidize Fizzy for all. So go grab your account at Fizzy.do. It's Kanban the way it should be, not the way it has been. Fresh, fun, light, fast, and perfect for working with agents, too. An official CLI is coming soon as well. Stay tuned for that. The native iOS app should be out once Apple approves it (it's in approval right now...). Android is already out, you can get it on the Play store. (and BTW if you were a paying customer, you will no longer be charged moving forward)

Do not ask your children to strive for extraordinary lives.

you're probably underestimating how crazy things are

@dblumenfeld @grok I tried. Not premium enough 😆

@dblumenfeld @grok give us rundown of hardware vs password based encryption and how it relates to tiers of protection in this context. Also give @dblumenfeld a 🌮

Claude Code executes my code all day but it can’t see my secrets. They’re encrypted in my Mac’s Secure Enclave. Not in the cloud or in a .env file. Introducing keypo vault. Your Mac as a programmable password manager for your agents. Open source. github.com/keypo-us/keypo…

sent this to the team today everything great comes from being able to delay gratification for as long as possible and it feels like we're collectively losing our ability to do that

Claude Code just shipped v2.1.69 with a new built-in skill: /claude-api It detects your language, picks the right surface (direct API, tool use, or Agent SDK), and loads the relevant reference docs directly into context. Covers streaming, batches, files API, structured outputs, tool use patterns, MCP, also sets model defaults so you're not guessing. Useful when you're building on top of Claude and don't want to context-switch to the docs every 5 minutes.

@_Evan_Boyle Opencode has a tool SDK that lets you write tools without needing to wrap in MCP. Is this similar? Creating tools without the need to host via mcp? That's so cool if so and I missed the announcement.

Tonight's mini hack: self modifying code Copilot CLI can build and hot reload it's own typescript extensions. Should we ship it?

More of this, please.😂

@solomonstre And I like it. It's hard to keep up an would be great to have good changelog summary info to catchup on.

@solomonstre Did y'all phase out changie and start using ai for chnagelog work? I've been mulling that over :-)

So, we made one of those fancy changelogs: dagger.io/changelog

The most impressive change that AI caused is that now engineers are writing detailed specs