today one of our media editing suppliers asked for consent to create a deepfake of my voice and sample data reading a script with all the major phonemes and I went nope nope nope nope nope I am okay with finetuning voice recognition, but not letting a corporation literally put words in my mouth.
every single existential risk I can think of is *humans* applying AI to bad ends, and not *AI* going rampant. Yes, a bioweapon would be bad. but it's more than likely a terrorist group that deploys it for psychopath reasons, not an AI trying to hold humanity hostage in exchange for more compute
I've been accepted into @anthropic.com's credits for open source developers programme, on the basis of my arm64 compatibility and optimisation work. I'm very excited to not have to pay for my OSS development tokens any more!
You are not immune to falling for AI image slop, I've seen like 3 viral images go around my timeline from people that I know wouldn't deliberately reshare AI slop. :( :(
Huh, I realised the switching friction and moat between harnesses probably doesn't exist any more. You can just have your new harness port your old harness's data over as the first thing it does. If it's good enough to reinstall an Ubuntu VM and port the data it's good enough to port LLM configs.
oh hey, genuinely, thank you @aaron.bsky.team, just had someone sealioning in my mentions who doesn't follow me, I don't follow them. in a conversation about AI use and bugs in popular AI tools, they came to tell me to just not use AI. notification popped up, disappeared, it's now a hidden reply.
3 fully LLM generated pitches for your startup to my inbox *in a single week* without any indication of interest from my part or encouragement to continue = instant block.
Anyone else notice that LLMs have stopped inventing fictitious story point / T shirt size estimates? They finally actually estimate post-LLM work units instead of pre. I used to have LLMs say "well that'll take two weeks, we'd better stop here" and I'd say "well try anyways" and done in 30 min
gaaaaaaah I've hit the point of pressing my yubikey or touchid becoming the blocking factor for my LLM productivity, and I'm sometimes _barely_ reviewing the requests before mechanically approving, except sometimes it does ask to push something I don't want pushed, so the gate is helpful but aaaaa
Looks like I'll close out this month closer to $1,500 rather than $5,000 in Anthropic model spend on my work account. A majority of that savings comes from shifting OSS dev to an Anthropic-paid account, but also I did get significant benefits from moving to Opus 5.5 medium from Sonnet 5 high.
