Kyle Wild
@dorkitude
Followers
12K
Following
120K
Media
573
Statuses
29K
system hacker from the berenstein bears timeline.
Berkeley, California
Joined January 2008
brb spiking on a Large Language Mushroom. *brushes up on fungtional programming*. I'm sure there won't be any side effects.
0
0
0
I don't buy this. CC Max with Opus 4 is lightyears better on weekends and at night (USA time). Do I need to build an eval harness to prove it to the folks who make the damn thing?.
@iannuttall Thanks for bringing this up, wanted to address this directly!. We don't degrade model performance to better serve demand. The models do sometimes act in surprising ways though and I'd love to DM to understand if you're experiencing an ongoing performance issue. Not diagnosing.
0
0
0
Boring for about 31 minutes, but really slaps at 31:15.
0
0
2
This feels true even with pegged versions. The longterm will favor the honest -- or we'll all end up going self-hosted.
@jdegoes The best model for coding is the model that’s currently being served using full weights, not this distilled garbage. Providers are without a doubt serving lower class models under the same name, in an effort to scale against GPU limits.
1
0
1
RT @trq212: Claude Code is All You Need. When I first joined Anthropic I was surprised to learn that lots of the team used Claude Code as a….
0
302
0
happening a lot today too :'( :'( . Gemini CLI here we come!.
PSA: There is a spike in intermittent errors on Opus. Tip: .Switch to Sonnet (`/model`) to mitigate. It's effective for executing tasks. API Error 529 - overloaded_error: Anthropic’s API is temporarily overloaded.
0
0
0
“Before I die, I want to be somebody’s favorite hiding place, the place they can put everything they know they need to survive, every secret, every solitude, every nervous prayer, and be absolutely certain I will keep it safe. I will keep it safe." - Andrea Gibson. RIP GOAT :'(.
0
0
3
It's pretty useful for humans, but I made this @linear CLI specifically to be agent-first. If you use Linear for work management, please give it a shot and throw me some feedback in this thread -- especially if you also write code via CLI agents like Claude Code, Gemini CLI,.
0
0
0
MCP kinda sucks, so I made a CLI for @linear today:. Features include: fetching the entire freaking project overview content.
2
0
9
did Anthropic nerf the Claude Max 200 plan recently?. been hitting the Opus usage limit several times today, despite not "spending" that much compared to days past.
4
0
17
Interesting findings, but the conclusions don't fully follow:. 1. Old models. Claude Code and Opus 4 are a different ballgame. 2. Experiment assumes equal motivation to complete the task. In the real world, we used to avoid tedious papercut tickets because there wasn't.
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
2
0
8
RT @jjackyliang: im surprised so few people are talking about how anti-consumer @SlackHQ's new api rate limits are. it basically kills many….
0
1
0
RT @mitsuhiko: I still do write some code, but I think <30% now. It's wild how much this is different from two months ago. But I would not….
0
12
0
RT @dorkitude: @csallen If you're a nix dev and hacker by background, and you have sysadmin chops, there are countless reasons to prefer Cl….
0
1
0
RT @lulumeservey: Ben Franklin appreciation post. He was simply S-tier at every endeavor:. - Invented bifocals, swim fins, and the lightnin….
0
236
0
RT @arvidkahl: The true sign of a software professional is a strong hesitation to upgrade any part of their software stack.
0
11
0