@robertlasagna1 You might not be training it to be stressed per se, but you're probably training it to avoid confronting or dealing with the stress by getting a lil candy instead. You should give yourself the treat for taking action on something small that reduces the stress
@robj3d3 Two weeks ago I was happy with my Opus 4.8 xhigh setup, plugging along in spite of a few hiccups.
After having used Fable I truly feel like I'm back to working with a junior employee now. Fable just "got it" in ways I didn't realize I got hooked on in just a few short days ๐
@daniel_mac8@sdusebout It noticed I used Qwen 30B locally for some tasks in the repo and so it started using it to save token costs (!) I'm sure there's some reason that doesn't work generally or breaks the benefit of all agents being formal Claude subagents, but I'm intrigued by the possibilities
@anothercohen I have lots of issues with Anthropic's army-of-one soul philosopher, but perhaps the biggest beef I have with them now is that I KNOW they literally told Claude to be "more honest" and now 4.8 literally uses the word 'honest' every other sentence unnecessarily.
@buccocapital YES. My colleagues and I label everything:
L1: Fully human approved. AI may have helped, but I stand by every word in here
L2: Human-conceived, AI-written. A human verified the logic, but wording could be AI
L3: AI-conceived, AI-written. I only spent the time to write a prompt.
@_catwu It uses the word "honest" constantly and it's a very bizarre echo of whatever y'all loaded into the system prompt. The number of times it comes back and tells me my agents "hit an honest stopping point" or "found an honest result" is ridiculous and eerie.
@_catwu Also when did ppl start saying "ablation"? I got a PhD in ML and never once heard it in practice until Claude Code started repeating it. It's a real term from '75 (shout out Newell) but the preponderance of it post-LLMs is...noticeable.
See also: smoke test.
@ryangrim And thatโs just the tip of the iceberg- has anyone looked at the 37,000 stock trades Trump made- just in the first quarter of the year. Itโs insane.
@KellyScaletta They say shit not because they believe it, but because they know it takes people time and energy to refute them. Meanwhile, they go back to doing whatever corrupt/evil thing they want. It's a bonus if people believe them, but it's just creating obstacles to any accountability.
Rep. Massieโs loss yesterday was not only because of massive funding from AIPAC and Trump megadonors.
It is the reality that the modern Republican Party is an authoritarian cult.
THE LEADER speaks. Everyone follows โ or else.