Brandon Wang

Claude .md ? he’s a doctor too?

@kais_rad I agree — on iOS it drives me crazy

@bradr on the web it is... this is like saying why don't we all just use the un-styled html buttons

it's unclear to me whether it's an application's responsibility to override poor design on the platform-level here, Linear uses a native Windows dropdown which is... ugly. in fact so many people override Windows native dropdowns that a lot of people can't even tell what a native dropdown is anymore. is that good?

BREAKING: a16z is now including 50g of PROTEIN in each term sheet

@sashaorloff congrats Sasha!

VCs told us accountants will never leave QuickBooks. But here we are. Active accounting firms on Puzzle are growing like crazy. Turns out accountants do switch. When you build something that respects their time. When the company behind it is building for them instead of against them. I'm proud of this. But this graph represents something bigger than us. The "accountants won't move" thesis is dead. They're moving. And the firms that moved first are telling other firms. Word of mouth. The same way accountants have always discovered software. Except now they're not recommending QuickBooks. Accountants don’t want to be the first, but they don't want to be the last either.

Part of the trend towards "ambient AI" is to address the core challenge that regular folks do not interact with an unbounded "ask the AI anything" textbox the way that developers expect Users tend to engage with these features at the extremes-- asking irrelevant questions (to learn nothing about the product's capabilities) or asking an overly difficult question which then lets the user down

Yes hascodexratelimitreset.today Whoever made this, you made me laugh.

Mark Zuckerberg acquiring AI companies

@bykahlil @naturalpay congrats on the launch guys!

210 days ago, @naturalpay was just a one-pager and a memo. Today, we’re coming out of stealth. Natural is the agentic payments platform powering frictionless money movement between agents, businesses, and consumers. Wallets. Payments. Ledgering. Routing. Identity. Compliance. Credit. Observability. Risk. Everything needed to move money. Engineered for agents and designed for humans. These primitives give you the ability to transact without becoming a payments expert or stitching together a dozen fragmented tools. Huge thanks to our team of 10 (soon to be 25), our early investors, and the supporters who believed in this vision from the start. If you want to help shape how money moves over the coming decades, we’re hiring. And if you’re building agents you should probably be moving money too. Reach out. Read more about Natural and the products we’re launching in the blog below. natural.co/blog/introduci…

imagine if Dario came out on stage and was like “this is my brother, Duigi” and another Dario came out from sidestage with a different color scheme

@rtwlz You gotta filter out the grocery items (they require $25 minimum) :) walzr.com/postal-arbitra…

With the advent of cron jobs for LLMs across consumer (OpenClaw, etc) and enterprise, has anyone seen error rates or latency spike around the hour/half hour marks?

Great move from Google-- clever live discovery of API endpoints as well so the CLI isn't stale

new york state could make chatbots liable when it offers advice across a broad category of categories-- not just medical/legal, but physical therapy, mental health, architecture, social work, etc folding-sky.com/blog/ny-senate…

side-stepping a joke about why ChatGPT now turns a blind eye to long-distance archery trajectory calculations right after OpenAI signed a contract with the department of defense... ChatGPT has a broader pattern where it flags a problematic request, asks the user to "confirm" it's fine, then happily proceeds. it's an "i really shouldn't tell you this, but i left the key on the counter and i'll look the other way" kind of thing. this goes beyond showing users where the limits are. it often provides a yes/no way to circumvent the safety filter. worse, the back-and-forth itself often satisfies downstream filters — the model thinks "i raised the concern and resolved it, we're good." some examples with real harm potential (eg. deepfakes): - chatgpt draws the red line, then offers to generate something right at it - if a request is problematic in multiple ways, it walks back some dimensions but not all, then proceeds - chatgpt "asks to be deceived" — eg. asking the user to confirm a photo is of someone over 18, then proceeding on their word my guess is the model is fighting its own safety filters, its mandate to be helpful, and its rl-trained propensity to ask for continuations. as models lean more toward helpful over safe, this will get worse.

“Improved UI for banned users” will be the title of the book on how Google lost the AI era.

"we love seeing boundary-pushing in this community" really? the revealed preferences seem... very revealed, if not spelled out explicitly in the literal previous sentence

@lanceyyan @robotaxi @Tesla interestingly, every maps platform absolutely refuses to make that left turn so you might need more than a day at Tesla, you might need a day each at Google Maps and Apple Maps lol

@bradr @robotaxi @Tesla the left turn is available, google maps outdated