Quarq retweeted
anthropic's backers are reportedly charging a 3–5% fee on top of the IPO price. even a $1.4T company has a delivery fee.
should've asked @askbro_io
As agent traffic increases, many businesses that rely on ads and monetise user clicks are going to die.
I don't think ads are going away, rather, they're going to change.
Running ads is renting real estate on the app's screen. This real estate is capped by the bandwidth of the human eye (which can only see about 10 products at once).
Whereas agents can explore 1000s of products before curating 3 suggestions for the human.
Ads in the future will be about being part of those 1000 products the agent explores.
Move over, B2B and B2C - the B2A economy is here
Agents are the primary users of the Internet, and soon will be the primary customers for every product
Our thesis @a16z on the massive opportunity here 👇 (cc @venturetwins)
most consumers are not really interested in email management, note-taking, or reminders
shopping and travel are the use-cases they're interested it
and a chat interface is not the right one for those
Quarq retweeted
production budget for @askbro_io's latest ad:
moustache: ₹40
balaclava: borrowed
handcuffs: a gamcha
cop: unpaid (@hrsh_ok)
criminal: unpaid, still tied up
meta says 10M+ people set up muse in a month and a half, and the #1 use case is shopping. more than a quarter of users buy through it, average purchase over $200. stripe had to raise credit limits in week one.
jefferies found about a third of the sites it tested block muse. the agent is ready, the websites aren't
"walmart's returns apply even when the agent doesn't act as intended." people will let agents shop wherever mistakes are cheapest to undo
As Amazon blocks AI agents from shopping on its site, Walmart is taking the opposite approach, introducing new technology that lets AI agents make purchases on customers' behalf, with controls over spending and transaction approvals.
spr.ly/6019BHXyoW
the new checkout agent we're building at @askbro_io doesn't pay on its own. it runs the flow, and when it hits payment or gets stuck it pings you: pay now, take over, or answer its question.
people don't want an agent spending without asking. they also don't want to sit and watch it work
the number from this i keep coming back to: messages per user on @askbro_io dropped 25%+ the same week cart adds almost doubled.
for a consumer agent, sometimes a long conversation means it failed
US banks say they can't tell whether an AI agent in your account was sent by you or went rogue. so morgan stanley, jpmorgan and citi are just blocking them.
india might have dodged most of this by accident. an agent can fill your cart, it can't type your UPI pin. every payment still needs you for one tap.
the US is now trying to build that layer after the fact
"instamart losses peaked in the september quarter"
said in the same breath as stepping up investment. both can only be true if orders per dark store grow faster than the store count
Swiggy Ltd expects its quick-commerce business, Instamart, to have peaked its losses in the September quarter, as it steps up investments to accelerate growth amid intensifying competition.
(@sowmyaram_ reports)
livemint.com/companies/compa…
every service in india still wants you to download its app. an agent that can open those apps for you turns most of them into a fulfillment layer.
that's where we're taking @askbro_io next
phoneuse is racking real iphones with real SIMs, tapped by a physical stylus, so AI agents can use apps that block emulators. their own example is "reorder my usual from zomato".
in india a lot of what people want done lives in apps, not websites. the agents that work here will need both
agents just find what humans left lying around, way faster than humans do
The agents got in through credentials and connection settings left exposed in app configurations, public code repositories, and overly permissive cloud roles.
spr.ly/6016BHMGon
claude code now gzips every turn so it reaches the model faster. grok got faster this week too. no new model, just less waiting.
speed is the most underrated feature in AI right now. most people will pick the agent that answers in 3 seconds over the one that's slightly smarter in 20

















