Eric Winer

curly > waffle > wedge > standard > crinkle > shoestring.

I love all these pitches for AI tools and harnesses that claim "productivity went up by 30%!" Like, for 50 years people have struggled to quantify software engineering output, but now, we've somehow solved that problem, come up with a single perfect productivity metric, *and* it turns out that every tech startup, now matter how small, is measuring their 𝘗𝘙𝘖𝘋𝘜𝘊𝘛𝘐𝘝𝘐𝘛𝘠, and has been measuring it for long enough to get a baseline, and is correlating it to adoption of your LLM harness! Wow! Or maybe Claude wrote your X post and made the whole thing up.

i know, i know, but do you have to taunt me with this non-dismissible banner?

alternatively: am i in danger? are they trying to warn me? what does Olivia from Geico know that i don't?

why do all my customer service calls end with them saying "stay safe"? is that a covid corporatespeak holdover, or is it like a common way of saying goodbye to people these days?

@andrewfullerxyz @Gemini this is chart gore. can you pin these axes please?

this tweet inspired me to read Grant's memoirs. thanks! also vibecoded a map of his travels & battles at grant.winer.ly, which has been especially helpful for the pre-war bits

@Langerius q: @Langerius did you handwrite this tweet? or was it openclaw, or maybe ironclaw? i appreciate the engagement regardless, just trying to track a pattern of behavior back to its source.

@eric_winer They almost earned that click

really interesting phishing attempt: they sent me a bunch of phishing emails, then called and did a legitimately decent Google account security checkup! even asked me to delete those emails. then 30 minutes later they called again and finally tried to get me to click a fake link

my single biggest production risk is @natebirdman's physical wellbeing. Nate, please tell me you don't own a motorcycle.

huh, i think it's actually been years since i last heard 'mr brightside'

@daylightco i would love to play with this

AI demos all try too hard to be fancy and impressive We wanna just focus on it being the most seamless, simple, & friction free It turns out it helps to have both the OS and hardware to pull off super buttery smooth AI UX We don’t even call it AI We just call it mAgic Ink Only on daylight. Comment below if you want to beta test.

seems like they've added some CYA crap to their usage policies, saying it's prohibited for you to use ChatGPT for legal advice without consulting a lawyer.  but i don't see evidence (yet) that they're actually changing or restricting model outputs.

is anyone else's battery life tanking after installing the iOS Vista update?

@guillewrotethis well there's a few examples of doing that in a NEAR contract: * github.com/peersyst/fast-… * emailbtc.net is there any value in doing the oauth verification in a shade agent instead?

Curious if anyone has built a Shade Agent that acts as an intermediary with the MPC contract, handling social authentication Basically, one could create a shade agent that checks an oauth token, and allow the user to make `sign` calls when authenticated

hey it's time for my monthly panic of "oh crap did i accidentally rack up a bunch of charges to my personal AWS account"

yay, more shrinkflation, how exciting.

Used to be I had 1Password and Authy. Totally manageable security. Now I've got passkeys littered across iCloud, Chrome, 1Password, TouchID, Yubikey, god knows where else. I know this is a Me Problem, but my online security has arguably gotten *worse* from the rise of passkeys.

Story time: Back in 2012, I went on a (third? maybe fourth?) date with a girl to a jazz bar that had lots of fun games. We had bonded over being competitive, and thought we could play some ping pong or foosball together. She asked to play Scrabble – but it turns out that makes it tough to have a conversation, since it involves alternating periods of intense concentration. I'm no Scrabble wiz, but I love using the two-letter combos to put words next to each other. So I kept doing that, and I remember my date getting pretty frustrated. Soon we ended up with the board below. At this point, we couldn't find any other possible moves to make ANYWHERE, regardless of the available letters. The game was stuck. And so was the date. Pretty sure we never talked again. I didn't remember taking a picture but it just popped up in my Google Photos. Sorry for the 2012-Android-in-a-basement-bar photo quality. If you're out there, fellow former NYC single, I hope you're well, I'm curious for your side of the story, and gg.

@near_ai @Google Ultimately, neither protocol has any adoption to speak of yet. We're still waiting for (and building at @near_ai!) the killer apps that will drive demand for agent-to-agent transactions. Once that happens, either protocol will work great, and work even better with $NEAR. (4/4)

@near_ai @Google Payments protocols: * AP2 is more fleshed out; AITP-01 Payments still has some loose ends * AITP's referral/affiliate revenue model is vital once you have agent-to-agent-to-agent communication chains (3/4)

Getting a lot of questions about @near_ai's AITP vs @Google's A2A/AP2: * Pretty similar in vision, goals, scope * Both move us towards user-owned AI; glad to see @Google making positive moves * AITP was there first; who cares? (1/4)