Luke Patterson

software shipper, technical founder, cat dad x3, musician

[Luke Patterson @lukenneth_]

@JoshuaSWarren @wesbos @syntaxfm throw a dawg a bone will ya?

[Luke Patterson @lukenneth_]

@kentcdodds I registered for https://t.co/nqdD3ehy06 but I'm growing concered that the materials will be outdated by the time it launches. At the current rate of progression, I imagine any tips outside of fundamental engineering practices won't be advisable anymore.

i.e. Who's to say that /skills won't be replaced by a more global pattern once we realize that invoking skills are a thing of the past?

[Luke Patterson @lukenneth_]

I keep seeing new startups that are made obsolete moments after their launch trailer by a frontier AI lab's latest update.

I actively avoid the temptation to build an AI startup and instead build a 'regular' startup super-charged by the advent of AI tooling.

The fundamentals of business don't change.

[Luke Patterson @lukenneth_]

@mattpocockuk at what point will we transfer skills into a global pattern like system prompts? strategizing what skills to invoke is a skill itself.

[Luke Patterson @lukenneth_]

@mattpocockuk @icanvardar the point of the patterns was always to have a shared understanding and form of communication with other engineers. makes no difference whether that engineer is a human or agent!

[Luke Patterson @lukenneth_]

installed omarchy last night, learning to use it feels like a new hobby, but im trusting the process. short term pain for long term gains. sped up rapidly by the open source community! i think this is the start of something truly magical and the world has no idea

[Luke Patterson @lukenneth_]

@JoshuaSWarren you're killing it! love what you're doing, thanks for all the hard work you're putting in for the community

[Luke Patterson @lukenneth_]

@poteto @lucasmeijer how does that compare to t3 code? using t3 right now across 3 machines. still need to watch the grok bot 101 recording i signed up for

[Luke Patterson @lukenneth_]

@theo I've switched to T3 code last week from the codex & claude desktop apps, which has been amazing with T3 connect across my many machines. and now I'm experimenting with grok bots. how would you describe the different mental models for coding with t3 vs grok bot?

[Luke Patterson @lukenneth_]

@theo Thanks for this! Felt the same thing and just bought 3 things off your list.

[Luke Patterson @lukenneth_]

@theo I currently use @jumpdesktop. software solution I know, which is fine for my use cases. any reason I might wanna switch? Is the latency any better?

[Luke Patterson @lukenneth_]

@_MaxBlade Thanks! I was just looking for something like this! Also coming from Mac, do you know if i can get my 2FA text messages to my phone appear in the text field like iMessage does?

[Luke Patterson @lukenneth_]

@mattshumer_ it's sad to talk to my 9-5 friends who have interest in AI but are limited by their company's token usage. not enough ambition for them to get personal subs and too much bureaucracy in corporate to put them on the frontier

[Luke Patterson @lukenneth_]

@AlemTuzlak I ship this much with neither of your prerequisites. But I also have over a decade of coding experience

[Luke Patterson @lukenneth_]

Was just thinking about something like this. And tailscale can enable a full stack app across a few machines on different networks (home & office)

Useful for when building for iOS and your Mac is located elsewhere

[Joshua Warren @JoshuaSWarren]

Just got a Mac on macOS and a Mac on Omarchy Linux to run one AI model together, each GPU taking half.

Same answer you'd get from one machine.

Now I'm building the Omarchy app that finds every computer on your network and turns them into one giant AI cluster, no matter the OS.

[Luke Patterson @lukenneth_]

brb im uber eatsing an xps 13 to install @OmarchyLinux @dhh

[Luke Patterson @lukenneth_]

@dhh My shed has an M4 Max Mac Studio that I remote into from a 2020 MacBook Air. But trying to look into a Linux box for more parallel work. Prefer hosting my own instead of cloud agents, any suggestions?

[Luke Patterson @lukenneth_]

Need to update my agents to prefer visual artifacts over text explanations. A picture is worth a thousand words, and they can type far faster than I can read.

Perhaps one way to remove the bottleneck for human review is to slow down the path to human review readiness.

[Luke Patterson @lukenneth_]

We must completely rethink everything about the shipping of software to be agent native, but with human oversight and orchestration.

I’ve created a new rule for agents that I should not be bothered with trivial changes and all PRs need a summary card with the least amount of information required for human understanding.

They type faster than I can read.

[Luke Patterson @lukenneth_]

OpenAI’s Dot is phenomenally good at helping me overthink everything and move at about 60% speed