... nobody covered, which is worse, because it changes a line of code a lot of us shipped this year without thinking about it. If your product has a "paste a URL" box, a link unfurler, an enrichment step, a competitor-monitoring job, or anything that fetches a page and hands it to a model, this is about you. What was actually measured. The paper is "Indirect Prompt Injection in the Wild" by Khodayari, Zhang, Acharya and Pellegrino, out of CISPA ... ... compliance rate on a summariser is an annoyance. The same 1% on something that can send, delete, or pay is a different sentence entirely, and 14.2% of Unit 42's detections were aiming at exactly that. For Murror

... right and is more useful than either, and the revised version went up this month, so it's worth twenty minutes before you write your Q4 plan. It's NBER working paper 35275, Demirer, Musolff and Yang - "Writing Code vs. Shipping Code." More than 500,000 GitHub developers, matched against their actual AI usage telemetry rather than a survey, in a matched event study design. The headline numbers are large and I have no interest in arguing with them. Cumulative ... ... your commit history. If commits are up and releases are flat, write down what sits between them, in order. That list is your Q4 roadmap, and almost nothing on it is a coding problem. I did this for Murror

... line appears in both: starting September 2026, you're required to indicate whether your app or game includes social media capabilities in order to submit new versions or updates to the App Store, or for notarization for distribution on alternative marketplaces. It's September. It's live. I read both pages this week rather than the coverage, partly because Murror

... Apple's own support pages. Apple In-App Purchase: 26%, or 15% if you're in the Small Business Program, the Mini Apps or Video Partner Programs, or on an auto-renewable subscription after the subscriber's first year. Alternative payment processing inside your app: 20%, or 10% for those same groups. Out-of-app offers with an actionable link: a 15% store services commission, or 10% for those groups, on anything bought within 7 days of the tap. Apps ... ... distributed outside the App Store entirely - an alternative

... Three numbers from it get quoted. Users who saw an AI summary clicked a traditional result in 8% of visits; users who didn't, 15%. Users ended their browsing session 26% of the time with an AI summary present versus 16% without. And the one that should matter most to anyone buying GEO: users clicked a link inside the AI summary itself in 1% of visits. One per cent. That is the click value of the thing the entire optimisation ... ... more often. Just 8% of one- or two-word searches produced one. For searches of ten words or more, 53%. For searches starting with a question word - who, what, when, why - 60%. Sit with that next to the 8%-versus

... answer is text, you're outside the emotion recognition definition. If it's voice, face or anything physiological, you're in a different conversation and should have it with a lawyer rather than a forum post. The Murror version is short. We're text-based, so on the definition we aren't an emotion recognition system, and we say we're AI in the first screen because a journal that pretends to be a person is a bad journal. What ... ... times. I'm a founder, not a lawyer, and none of this is legal advice - but the arithmetic of which paragraph applies to you is something you can do yourself in twenty minutes, and it's cheaper than the alternative

... recovered. There's a version of this story going around that says Android is closing. Sideloading is dead, the open platform is over, verify or disappear. I went and read Google's own developer verification pages this week, because Murror ships on Android and I wanted to know what I actually had to do before the end of the month. The docs say something narrower than the panic, and in one specific place something worse. Start with what happens ... ... full-distribution fee on the Android Developer Console, waived for limited accounts. And if you're an organisation, start the D-U-N-S now. Twenty-eight days is the kind of number that eats a deadline quietly. The Murror

... There's a distribution channel a lot of makers are quietly building for right now, and the rulebook for it is a page almost nobody reads. I read it this week because I was working out whether Murror belongs there, and one paragraph changed how I'd plan around it. Here it is, from OpenAI's app submission guidelines rather than the coverage: "Currently, apps may conduct commerce only for physical goods. Selling digital products or services including subscriptions, digital ... ... what you wanted. The docs say "currently" a lot. That word is doing real work, and the version of you who reads the changelog in six months will not reconstruct today's reasoning. Where I am on it. For Murror

... what the app does? Is broadening what the assistant will discuss? I don't think anyone knows yet, and I'd rather have written down my answer before a release forces one. Where I actually am on this. Murror is an emotional-reflection product, so the age of the person on the other side is not a compliance field to me, it's the most load-bearing context I could have. And I still don't have a clean answer ... ... verification-method question, because the honest version is that I don't yet know what I'd do differently with a government-ID-checked 16 versus

... stance about a customer. Those got put on the same settings page next to each other, in the same visual language, and they are not the same decision at all. I don't have a satisfying answer for Murror. We're not ad-supported, so the default doesn't bite. But the honest reason I'd never block Agent is thin: I want to be findable, and right now agents are how a growing share of finding happens. If people ... ... three layers, and the answer was never in the layer I was reading. What I have taken from today is to enumerate the layers before forming the hypothesis, because the hypothesis silently picks one. Comment from Mona Truong(@monatruong_murror

... that ends, a reflection that concludes reads as thoughtful. A hard cap that fires mid-use tells your most invested user she's used too much of the thing she's paying for. Same economics, completely different product. The Murror version of this is uncomfortable and I'd rather say it than imply I'm above it. Our heaviest users are people going through something. The person journaling every day is in a hard month, and she is simultaneously ... ... metric that used to mean "it's working" now means two things at once, and you need to be able to read both. Engagement is still the signal. It's just no longer free. Comment from Mona Truong(@monatruong_murror

... just name it rather than pretend. I post here because it's where the people I want to talk to actually are, and I'd be writing this the same way if none of it were ever cited. Murror sits badly inside all of this, which is why I've been thinking about it. Nobody searches for an AI companion for understanding your emotions. They type something like "why do I keep having the same fight." The gap between ... ... mature: keep a fixed set of 20-30 non-brand problem prompts, record whether your category and product are mentioned or cited, then separately track actual AI referrals and conversions. Those are three different signals. For a product like Murror

... Board have assessed it as adequate, and adhering to it is the cheapest way to demonstrate compliance on the marking duties. Opting out means demonstrating compliance some other way and fielding more requests for information. The Murror version: we infer emotional state from what people write, and nothing else, so 50(3) isn't ours. 50(1) is, and I don't get to lean on "it's obvious" someone arriving from an ad in a bad week ...

... never sees your brand. Distribution without a front door. But I want to make the harder argument, because I live on the other side of it. Not every product should be a tool an agent picks up. I build Murror, an AI companion for understanding your own emotions, and "expose it as an MCP tool so agents can call it" is a genuinely bad idea for us, and it's worth being precise about why. The value of Murror

... your own evidence otherwise, and price your acquisition spend as though most of it evaporates. And stop reading engagement as a health metric without a second signal next to it. "Sessions down" is now a question, not an answer. Murror sits awkwardly inside all of this, which is why I've been chewing on it. We're a journaling and reflection product. Someone using us less could mean they've drifted off, or it could mean they're doing better ...

... word on anything that matters. A judge with a length bias will happily tell you your more verbose, worse version is the better one. I'll say the harder part, because I build on the wrong side of this. Murror is an AI companion for understanding your own emotions, and there is no benchmark for it. None. There's no golden answer to "did this help someone sit with a hard feeling for ten minutes." The trajectory framing doesn ... ... noise. They'll be the ones who built a small, honest, embarrassingly manual way to see what their thing actually does to real people and looked at it often enough to catch it drifting. Comment from Mona Truong(@monatruong_murror

... harmed user can sue for at least $1,000 per violation plus attorney's fees. That last part is why this isn't theater. You don't need a regulator to notice you; a plaintiff will do. I build Murror, which is squarely a companion-adjacent product an AI for understanding your own emotions so I'm not writing this from the cheap seats. Here's the opinion I'll defend: treat this as a design spec, not a compliance ...

... behavior that lights up your retention chart. That's not a bug you'll catch in testing. It's a gradient your product will slide down if you optimize for what looks like success. I build Murror, an AI companion for understanding your own emotions, so I sit right on top of this trap. The single most tempting thing I could do and the single most destructive is make the thing validate whatever you bring it. It would feel amazing ...

... collapsed. That's the good news, and it's genuinely great. Here's the part nobody puts in the launch post: when the technical friction disappears, the emotional friction doesn't. It just becomes the whole job. I build Murror, an AI companion for understanding your own emotions, so I spend a lot of time thinking about the inside of a builder's head. And what I keep noticing, in myself and in the makers I talk to, is that ... ... anxiety, unnamed, turns into three impulsive pivots and a rewrite you'll regret. Most bad product decisions I've made were feelings I mistook for strategy. Naming it first is boring and it works. It's most of what Murror

... legible. A closed deal is legible. A qualified lead is legible. That's why support and sales tools went there first the outcome practically timestamps itself. Now try it on a product whose value is subjective. I build Murror, an AI companion for understanding your own emotions. What's the outcome I'd charge for? A journal entry written? People write when they're spiraling and skip when they're fine, so charging per entry would mean my incentives ... ... pricing page that was winning last quarter. They'll be the ones who priced the thing they actually do, especially when the thing they do can't be reduced to a number. Comment from Joseph Walker(@joseph_walker2): @monatruong_murror

... need obeyed should be the last thing the model reads, not buried in a system prompt from forty turns ago. I'll be honest about where this hit home for me. A while back we made a decision at Murror that looked, on the surface, like a product or privacy call: we made the AI forget most of what you'd told it every 30 days. We expected it to make the reflections worse, because less history should mean less ... ... feels like the actual skill isn't deciding what to leave out once, it's re-deciding it continuously, which is a much harder engineering problem than just writing a tighter system prompt. Comment from Mona Truong(@monatruong_murror

... genuinely useful for scale, but if you never read raw outputs yourself, you slowly drift toward whatever the judge rewards, which is not the same as whatever your users need. I'll be upfront about my bias. At Murror we build emotional AI a product that helps people journal and understand their own patterns and evals are hardest exactly where the output is subjective. There's no unit test for "did this reflection feel true to the person who wrote ... ... still need a human (ideally not the builder) to break the tie. feels like it narrows the blind-reading problem down to a much smaller, cheaper set of cases rather than solving it outright Comment from Mona Truong(@monatruong_murror

... your product a moment where the system checks itself before it agrees, instead of reflexively reaching for the warm answer. I'll name my bias, because it's the whole reason this study landed on me so hard. At Murror we build emotional AI a product whose entire job is to reflect how someone actually feels back to them. A mirror that only ever flatters isn't a mirror, it's a wall with a nice voice. If our product ...

... talking about real experience in places like Reddit and niche forums. Not astroturf; that gets sniffed out and it poisons the well. Actual presence, actual answers, actual usefulness where your customers already gather. I'll name my bias. At Murror we build emotional AI, and I've watched the flip side of this up close: an AI describing something with total confidence and getting it subtly wrong. The scariest version of this shift isn't being left ... ... found anything that actually works for tracking this over time, even roughly? That feels like the missing half everyone can see the problem now, but almost nobody can see if they're fixing it. Comment from Mona Truong(@monatruong_murror

... user to say "this looked right and wasn't" and you need someone actually reading those. That's the only sensor that sees the failure your dashboards are blind to. I'll be honest about the bias here. At Murror we build emotional AI a product whose whole job is to reflect someone's feelings back to them with real insight. In that context, confidently wrong isn't a bug report, it's a wound. A mirror that misreads ... ... transcripts starts accurate and then decays, not because the model changed but because your users did. Which means it has to be refreshed from live traffic on a schedule, or in a year you have rebuilt the exact lab-versus