"yo carly create a state action dispatcher that alerts on the (sanitised) event-stream of the upstream dependences of anything in the fleet, then dispatch each to auto update fleet wide when the upstream lands any SemVer-change off primary release channel (and the beta stream/channels optional sub-settings also same dispatched auto update respectively please)...
Double check each flake gets checked for complexity reduction if something meaningful changes in the upstream as well please. The full fleet sync has to stay within the flame profile limits we set from event to live usage."
Berkle retweeted
If your codebase is structured like 💩, an agent will smear and replicate everywhere—all the bad patterns.
So it's always a good idea to lean into the latent space every now and then and use its computational ability to choose better names, module structure, file names, and directory structure than a human can do on their own.
I call this a janitorial Roomba task. Something like this is a great fit for a skill and can run as a background cron job or a GitHub Actions job off your trunk branch.
I also have seen many things, I have never seen that. I didn’t think we’d get there this fast either
The world is getting weirder and more amazing every week.
With accelerating progress (and ASI), we will have flying cars everywhere (real/proper flying cars, not drones pretending to be cars) by the 2030s.
The only thing that could prevent this, would be regulations.
But the technology will be there. With SpaceX tech, Tesla can do it.
Been working on some tern plugins for good dashboards and access to my fleet.
It gets wild when home system grows in complexity.
going for functionality first then nice visual fidelity later!
but at least the animations are clean and nice and its fully interactive which
tbh its like making your own custom OS.
stencil.so/tern absolutely is the future
Hey @thsottiaux as you continue your 30days of Christmas marketing to save face for cutting the token budgets,
Could you please fix your models RLHF-fried prose generation?
I constantly have to say things like this after literally every artifact is delivered:
“That’s an alien AI artifact that is so cognitively burdensome to consume, to read the prose at all is, like, not worth using in its current state.”
And yes, I could write an extensive process to use decision models to assist in validating, rewriting, and correcting this prose automatically.
If that’s the idea, then we’ve really missed the boat here.
Sincerely - a dev bummed that openai isn’t was they used to be
Berkle retweeted
Here’s the source & tutorial. Play with it yourself!
chenglou.me/handwired/
github.com/chenglou/handwire…
hmm maybe some type of home solar smart command center plugin in tern? Slop so far.. but I'll use ugly functional until the models can 1shot good ux, or a stop being lazy and spend time and effort to max that out. only so many hours in a day
Ive gotta say as unconventional as using profane language is with models, when I go dad voice, shit gets done
"Why is the Ghostty skill still called the Ghostty skill if it's now only referring to tern and there's no more Ghostty?
Isn't that the fucking pure example of perpetuating tech debt and nasty fucking spaghetti code into our operation? AND the diametic opposiite of whats good for this system and what I specifically tasked you with.
What other fuckery is afoot in this PR? Please do not leave garbage trail in my repo, fix your shit"
in the spirit of learning from the greats
this mornin
with the following random prompt
(techniques gleaned from the likes of @yacineMTB and @elder_plinius and @doodlestein and others)
"Good job go ahead and merge and sync that, then immediatelly embark on an addtional triage or other out of the box solutions, new innovations in the fields of the types of bottle necks identified in the flame brake downs we can generate from the profiler in tern and debug tools we have.
Get those many many obvious still existing problems here. Human beings are horrible at writing code, so lets ultra bench max thruput with this. Id bet we can improve massively at leasst 10-15 different items easily, probably 40 or so. Get it to 10-15x faster across all parts of the core bottlencks in the lifecycle of this recap feature all the way thru. More reliable, faster, less code, more durable on all
all objectives."
Its like 'sleazey sales guy' trying to 'reverse-tom sawyer' a 'giga-chad-autistic-super-cosmonaut-sometimesmayberetarded developer
(opus/astra/argon/any of the like)
But it fucking works I gotta tell ya..
