Garrett

The world is built on spreadsheets. Its time they didn't suck > TermSheet The spreadsheet I have always wanted (The final feature is the real killer)

Dont you love it when the exact feature you need comes out on the same day you need it

@mitchellh Agree, I think they should take a page from @linear on their early 3rd party agent integrations. Their APIs and UX was *chef kiss* Though I think linear may be straying too far with now introducing their own coding agent & code reviews

Here’s what I’d do if I was in charge of GitHub, in order: 1. Establish a North Star plan around being critical infrastructure for agentic code lifecycles and determine a set of ways to measure that. 2. Fire everyone who works on or advocates for copilot and shut it down. It’s not about the people, Im sure theres many talented people, youre just working at the wrong company. 3. Buy Pierre and launch agentic repo hosting as the first agentic product. Repos would be separate from the legacy web product to start since they’re likely burdened with legacy cross product interactions. 4. Re-evaluate all product lines and initiatives against the new North Star. I suspect 50% get cut (to make room for different ones). The big idea is all agentic interactions should critically rely on GitHub APIs. Code review should be agentic but the labs should be building that into GH (not bolted in through GHA like today, real first class platform primitives). GH should absolutely launch an agent chat primitive, agent mailboxes are obviously good. Etc. GH should be a platform and not an agent itself. This is going to be very obviously lacking since I only have external ideas to work off of and have no idea how GitHub internals are working, what their KPIs are or what North Star they define, etc. But, with imperfect information, this is what I’d do.

Opus 1M context is still dumber than rocks after 100k tokens

@tannerlinsley CC for anything where cost is an issue Pi for anything quick or for personal agent stuff

Ghostty was fun, but time for something else. I still love opencode, too but with CC plans dead on it… I’m feeling lost. Full GUI? T3 Code? Opencode GUI? Warp? Back to cursor? Try CC again? Raw Codex? My 🧠 hurts and I just need to keep shipping.

@DavidKPiano @MaineFrameworks This man event sources

@MaineFrameworks Sounds nice, but where do you think state comes from? State is a lossy snapshot of all events

Events and state are dual, neither is fundamental

@thdxr SST was the only thing that made AWS manageable before

it took us several days to get our new AWS accounts properly provisioned with the right capacity limits to deploy our stuff it's crazy how things have regressed here

@mitchellh @charlieholtz @chaseonxcom @conductor_build just had this same experience with direnv trying to set up soloterm (cc: @aarondfrancis for constructive feeback. otherwise feels nice!)

Just downloaded it, feedback #1: execute commands in the context a shell initialization. I wrap all my projects up in direnv/envrc shells and only CLI-based agents get it right (because they're already in them). I can force it via AGENTS.md stuff ("wrap all commands in ") but its super clunky especially when going back to the CLI and everything is double wrapped.

I’ve been forcing myself to use the GUI agent apps the past week just to learn. One week isn’t enough experience yet, but I think they all miss the mark (so far) and I’m deeply fighting the urge to do one with my own taste. I’m not gonna do it. I’m not gonna do it.

@nizzyabi Its like instragram stories all over again

every major AI company is building the same thing

@homegymcoop I want a belt squat hooked up to one of those digital transducer things you showed off

This is the yet-to-be released Rogue Air Rhino. It combines a belt squat machine with the fan flywheel you’d normally find on a rower. Essentially, it provides unlimited resistance that instantly matches the user’s power. AKA, it will be brutal to use.

@BowTiedCrocodil I thought the software dev lifecycle was dead tho?

AI should never touch prod ever Version control all infrastructure code Only allow secure pipelines to execute the terraform Always run plan to see expected changes Always have code reviews of the terraform Have a pre prod environment to test changes Guys this is software development lifecycle 101

@BowTiedCrocodil @Walmart I have never had an online walmart deliver successfully delivered (i've tried ~5 times)

Tried @Walmart delivery today just to see if they could do day of stuff. First order said it would arrive in a few hours, go to checkout and pay, check in a few minutes, and says delivery by 10pm. Cancel it immediately. ❌ Try this again, but ensure delivery window is correct. Purchase and wait. Driver goes out for delivery, I can track them. I get distracted and check back, tracker now isn’t working. Wait a few more minutes, order canceled. ❌ Sick stuff @Walmart

@NathanFlurry Ill see what I can do!

we have a channel called #effect-wg in our Discord we have something that's 90% of the way there, but we haven't had time to dogfood it enough to be confident in shipping it if you have an hour or two, it'd be really helpful to pull our effect branch and give some feedback. we really want to ship this rivet.dev/discord

Just shipped: Rivet Workflows Actors already have state, queues, and fault tolerance. Step-by-step durable execution fits right in. 🪜 Sleep, join, race, rollback, human-in-the-loop 🔁 Infinite loops for long-lived workflows 🌐 React (useActor) 🔓 OSS + self-host (Apache 2.0)

@NightOwlCapital Built an ai agent for our biz connected to slack it does feel exactly like messaging one of our old contractors

Talking to employees on Slack feels no different than prompting AI Let that sink in

@thdxr Just waiting for someone to name one after a crawfish mud bugs are the tastiest but also the grossest

i like eating them but i've always found crustaceans really gross terrible time for me

@jaredpalmer yup x.com/pacificdaemon/…

It feels like current generation of coding agents are overfit to make code run (and render) successfully on every prompt. Too often, they emit “fallback” behavior and avoid breaking changes as a reward hack, sometimes not even telling you about it.

Its amazing how much faster and higher quality Claude is on the weekends

These claude subagents are getting a little out of hand - task - plan - task /accept - task - task (15mins later) ok now i can code

@MichaelArnaldi @thdxr @dhh opencompute

@thdxr @dhh Next up: build a datacenter

it's so weird watching people continuously pledge allegiance to multi billion dollar corporations and then experience heart break these are just things you use when they're useful save your heart for real people

@nnnnicholas Just tried to do this same thing last week. Still havent found a good solution

It remains too early to delegate format selection to an Agent Because it is actually very time cosuming to codify format CREATION into a script And so you're wasting time getting superficial research of formats you cant easily recreate.