Kenby

Kenby

@kenby.dev

ASIC Engineer

208 followers429 following182 posts

It simply doesnt bother me that people are making low effort software I used windows for years

Guy who thinks creativity in swe is how you choose to implement your sorting algorithm

I would love to go back to people not really caring how software was made again

sometimes im disappointed when a sci-fi authors vision for the future revolves around their fetish but at the same time, cant blame them for being optimistic ig

also if it bothered me that much id have stopped reading them bc 90% of sci-fi is this

So is Hermes just terrible? I've been using it for a while and find it pretty unsatisfactory but been thinking its a skill issue

Tried just not looking at the code for a side project Look inside... agent was reimplementing an http server instead of pulling in a dependency I think I'm gonna keep looking at the code

Every time I have some kind of appointment I feel like a caveman plopped into society and if anyone finds out I dont know what im doing I Lose

My dream is to get lucky somehow, retire super early and get to build whatever I want for fun tbh

I thought semantic recall would be cool but I cannot solve figuring out what to remember :/ In that way I guess it is like human memory lmao

I think im going to do one cheap llm pass to generate candidate memories another to reject candidate memories 🐴 so far ive just been doing the first pass and its been ~30% garbage

Used glm5.3 flash while it was free for a lot of stuff and I think the small model bug has bitten me For a lot of stuff I do not need a model better

How do I stop agents from building backwards compatibility with things implemented 4 minutes ago on a feature branch

Kenby reposted

deepmind is the only major US AI lab that has a serious StarCraft 2 bot and they shelved the project. absolutely disgraceful state of affairs

I want to write more (most?) code by hand but I have to move so fast at work that I feel like i cant ): I feel like im being deskilled in some ways. Maybe upskilled in some others at least. Hard to say.

Kenby reposted

"Oh thank God, finally done with Evil Computer Time, now I can enjoy Good Computer Time."

Sam Clemente@samclemente.me

Alright why is my brain so fucked that sitting on my couch and architecting software systems is a good way to unwind from a long day of...sitting at work...and architecting software systems

Kenby reposted

The idea that there's an easy abundance of free resources to learn to code is laughable once you've done what I've done and been a VP of research at a company that helped thousands of people break into tech careers. There's an abundance of *beginner tutorials*

Really struggling in a world where when people ask "who even wanted this" the answer is usually "me"

Kenby reposted

"Microsoft leaned on the Linux Foundation to make Linus Torvalds change his opinion about AI" is one of those things you can only say with confidence if you don't know anything about the current state of software engineering and also don't really know anything about Linus Torvalds or the LF.

Kenby reposted

one thing that I hate about ADHD is that I need to be hyper interested, extremely stressed, etc. to really get into something. no gears beyond locked in or kinda listless. this isn't a quirk chungus so wacky thing, it's a character flaw that makes my life worse.

Kenby reposted

have you seen the new frontier? it's on anthropic. grok has it. you can chat with it on openai with ads. glm has it. you can literally get it on kimi. kimi has it for you

Claude always bringing up a bug I told it to avoid once feels like how I always worried about quicksand as a kid

What providers are people using for glm 5.2? Looking to finally start playing with it tonight

Kenby reposted

Kind of sad how jaded people have become to technical progress. Natural language interfaces for computers were a holy grail of artificial intelligence for decades but as soon as we actually solve it the goal posts move to "ai needs to be as smart as humans to be useful at all".

Kenby reposted

I'm a social democrat that is in favor of technological advancement and limited intellectual property protections This, to me, feels like the only stance that is logically coherent for a social democrat, but I increasingly feel like an outlier

Kenby reposted

Guys, the “stop hiring humans” ads are bait. They’re designed to be angrily shared on social media to get them in front of as many eyeballs as possible. You can play a part in helping this strategy be less effective, by ignoring it

Kenby reposted

the more I look at social media the more convinced I become that the way many people use hyperbole is functionally equivalent to lying and is almost as bad for themselves and others as lying is

Is everyone else just copy-pasting simple errors to claude or are you guys going about it in a smarter/more directed way? Or I guess just setting your harness up so that it can iterate on them without input is best huh

I could just be coping about being too lazy to set up nice feedback loops for claude but I do like being in the mix a bit for scanning through errors and determining whats really relevant

Spent ~$80 on tokens to mess around with Fable last night on designing an MMO trading strategy planner. Im very impressed by the quality of the output but its just way too expensive. I would certainly never spend that again preferring to opt for opus and a hands-on approach

Kenby reposted

yea you should come hang out. i got the green Perimeter over at my place

Shodan Safari@shodansafari.lol

ASN: AS7922 Location: Plymouth, US Added: 2026-06-08T01:34 #shodansafari #infosec

Kenby reposted

a lotta yall still dont get it glasswing subscribers can use multiple mythos previews on a single agent so if you have 1 opus and 3 max subs you can create 3 new opus subagents Tonight's fable mint event is essentially a minting event for both Fable 5 and Mythos 5

This is crazy to read as someone who's whole job is verifying code written by people more expert than me

Ok, let me state one step more formally my problem with AI code. You cannot trust it to be more expert than you because you cannot check and verify what it created. It's beyond you. You cannot be responsible for it, nor can the AI. It can ONLY be used as a toy or subset of your abilities.

I honestly thought i was neurotypical for a long time but working in an office has convinced me I should get evaluated bro I cannot read these social queues

Kenby reposted

at the gym one teenage boy is comforting another through a break up while they do pull ups together. says ‘bro don’t even worry about her what if i get some edibles for this weekend bro let’s just get blasted and play GTA.’ honestly very touching