Arthur Etchells

Very important lesson for Blue Cities here: Every change Lurie has made has been opposed by screaming, bullying activists -- ignoring them has been the path to policy success and extremely high popularity. @skaushik100 sfchronicle.com/election/artic…

Reading scores, 3rd graders to 8th graders, 2015 to 2025. This is a national tragedy.

We used to go to a special website, ask strangers for help with programming, and get humiliated in return

I am down one hell of a rabbit hole Only like **70** years ago, pharmacists didn't want patients to know the NAME OF THE DRUG they were prescribed so the pharmacist didn't have to explain anything to them. But it wasn't exactly their fault? Check out the 1952 APA code of "ethics": "The pharmacist does not discuss the therapeutic effects or composition of a prescription with a patient. When such questions are asked, he suggests that the qualified practitioner is the proper person with whom such matters should be discussed."

@dbreunig Public fine-tuning APIs, are you sure they're ending all fine-tuning?

OpenAI winding down fine tuning is an interesting development and one to watch. On one hand, model maximalists will argue the largest models keep getting better at more things, so the need to adjust the weights of them is less necessary. On the other hand, the big labs keep pushing their models to a handful of use cases while training their harness designs into the model, rendering them less generalized. There's an argument _this is fine_, because coding and reasoning abilities will solve most other problems. But what we end up with are models build for their own harnesses. @badlogicgames was wrestling with Claude in the OSS Pi harness this week, trying to wrangle out specific in-harness behaviors, with Claude fighting him every step of the way. If this continues, there's a world where 3rd party harnesses become less valuable when used with frontier lab models because the 1st party harness behavior is already _baked in_. And there's no longer a fine tuning escape hatch to generalize this behavior away. Will then frontier models resemble appliances, not general platforms? With their harness trained in and no ability to adjust it? This might make application building easier for some enterprises, but the trade off is lock in.

Think of the morticians, crematory operators and embalmers! The deaths must continue that the unions be appeased ☠️🪦

@patrickc Lift (why do wings actually work?)

Which are the most common everyday phenomena that we don't properly understand? Off the top of my head: • Lightning (how does it happen?) • Sleep; dreams (why do they exist?) • Glass (thermodynamics of formation) • Turbulence (when does it start?) • Morphogenesis (how does a creature know what should go where?) • Rain (it seems to start faster than models would predict) • Ice (dynamics of slipperiness) • Static electricity (which material will donate electrons?) • General anaesthetic. (And the mechanism of a lot of drugs, e.g. paracetamol.)

Satya’s discipline on discoverable communications is impressive.

@raveeshbhalla Amazing

Some time ago, when I screamed go fuck off to my google home, it decided that my name was, in fact, Go Fuck Off. It however used beep the F word. Now, post Gemini, it no longer does. Incredible turnaround for a company accused of being too woke with their AI 2 years ago

@morriceGavin @hschnedlitz @morriceGavin still using git butler? Curious what problems it solves for you

@hschnedlitz Have you tried Git Butler?

TIL about git-worktree. It's a command that allows you to work on multiple branches simultaneously. Basically, it creates a folder within your git repository that contains a new branch. Fantastic if you want AI to hack away while you continue working in the main branch.

As a California taxpayer, I find this radicalizing. Our state must break the one-party rule and restore prosperity to the Golden State.

true

This was fun to write

@bonchieredstate Incorrect.

@themandeepc Tbh I’m excited to talk to someone who hasn’t churned off brain trust. What is the reason we’re still talking about them from your POV?

@asketch Good luck writing a Braintrust clone with a Claude Code sub. There’s a hundred competitors in the LLM observability market, and they sell to the most pro-AI agent demographic, yet there’s a reason we’re still chatting about Braintrust.

Seldom-appreciated effect of Claude code Codex subscriptions: Permanent shift in perceptions of software value. $200/month for the all-building-everything-miracle-tool means you better offer a lot of value. Braintrust @ $250 / month is ngmi.

@themandeepc Impeccable branding and an effective sales team 💫

HACK ON ROBOTS AT SPC MAY 15-17! Build something that moves. Three tracks: manipulation, aerial, & locomotion. Mobile robots, drones, robotic arms, and 3D printers on site. Apps due May 12th. Sign up now! 🤖

@cremieuxrecueil Now do generics

To learn more, go check out my latest article update: cremieux.xyz/p/how-to-get-c…

A new study has just come out vetting the quality of gray market peptides! What did it show?🧵 First result: The median and the 25th/75th-percentile peptide vials tend to be filled correctly, with between ~95% and ~110% of the amount they say they contain:

Tryna understand how Google is fumbling so hard they're hiring someone else to... search 👏 the 👏 web.

@ankrgyl Appreciate what y'all are building - reflecting on three different companies I've talked with who churned off the $250/month offering. All vibed a solution specific to their needs and this value calculus was loud for them.

@asketch Hi! Braintrust is no longer $250/month. There's a free plan and you just Pay-as-you-go based on consumption from there. $250/month gets you more features (like RBAC) & support.