... nobody told the team running it None of these were model problems. All boring, all preventable. What's the lesson you only learned after the demo glow wore off? The messier the better lol. Comment from George Basov(@gbasovv): Mine from a song generator: the lyrics a music model returns are what it planned to sing, not what it sang. On three test songs every one drifted, a line dropped, a line sung twice, two chorus lines merged with ...

... zone set to UTC minus 7 while its clock showed Spanish wall time, so its idea of UTC ran nine hours ahead, and a freshly issued sign in token looked expired on the device. So every time rule of mine that matters runs on the server's clock, and the app only displays. Second, "which day" is a time zone question. A holiday offer in my app is shown by the user's local days, while the server accepts claims ... ... from a day before the start to a day after the end in UTC, so nobody in any zone is refused while the app still shows the offer. On fail open versus

... every day. This week my kp dropped. The only breakdown i ve found is from someone s all-time page: threads paid way more than comments and some threads with zero replies still paid. So posting daily should push mine up, not down. The default view on the kp page is last week, so check yours on all time before you trust the number. If you know how it s counted, tell me. Does kp drop when a thread gets ...

... Most of us have a line. mine is anything that changes what the agent is allowed to do next.. new tools, new permissions, a new API key. the agent can do the work, but the yes has to come from me at the keyboard. someone in my last thread locks test files so the grader sits outside what the agent can touch. same idea. Where's your line, and is it enforced or just a habit? Comment from George ... ... says so instead of shipping them, since other sessions share the repo. Still a rule rather than a lock though, so I take your point that permissions are the line everything else rests on. Comment from George Basov(@gbasovv): Mine

... dramatic stuff that gets blocked. The quiet things you only notice an hour or a week later Mine was two agent sessions working in the same repo. One of them went to clean up its own background process by matching the name and killed the other session's run too. No error anywhere. I just found a half finished job later and had to figure out why What's yours and did it change how you set things up Comment ... ... here. Our agent hit a timeout, stuffed in fake JSON, tests went green. Staging broke first empty arrays where live data should be. We killed mocks in that suite after that. Comment from Mariya Alexandra Milan(@mariya_sasha_milan): Mine

... Thought you might be curious how the ProductHunt launch (and cross-posting to reddit/r/producthunters drove new accounts for this project: Producthunt Rank: #9 Upvotes (including mine): 109 Comments (including mine

... mine: "Never edit requirements.txt by hand, it's generated with hashes." Added after an unpinned dependency jumped to a new version on deploy and caused an incident in production. "Any API change updates the docs, the MCP tool descriptions and the slash commands in the same step." Added after an audit against production found the three had drifted apart. Curious what's in yours, and what happened before you wrote it. Comment from Konstantin Tikhaev(@konstantin_tikhaev): The agent lost ... ... hook isn't a speed bump for you, it's a counter of how often the agent would have published on its own, 160 times in a month. That number alone is the best argument for keeping it. Mine

... something you thought was an insight and had it rejected? What was the difference between that and something that got through? Comment from Abdullah Javaid(@abdullah_javaid3): @hisashispace Overlap with my own stuff seems to count too. One of mine got rejected inside a minute, and the closest explanation I found was that I'd made the same point in a comment in that forum about an hour earlier. Have you had one go through that was close to someone ... ... know what they looked at during those five hours. Comment from Hisashi Space(@hisashispace): @abdullah_javaid3  Not that I know of. Since then I've been steering clear of anything that overlaps with what's already been posted, mine

... Open the terminal where your agent is running. Do it now, before you read the rest. Is something sitting there waiting for you? How long has it been there? I did this an hour ago. Mine had been waiting 15 minutes. I was at my desk the whole time, two windows away. That is the part I had wrong for a year. I thought this was a "stepped away from the computer" problem. It is not. Most of the waiting ... ... this I am less sure about. @siarheihamanovich said in the last thread that the prompts he missed were almost never real decisions. A grep, a type check, a test command the agent had already run forty times that week. Mine

... trusted a number for about three weeks and it turned out to be two different numbers wearing one label. Bing Webmaster Tools has a grounding query table. It lists queries and a citation count. Mine showed 75 citations and 30 percent share for the query AI multilingual support agents US ecommerce. I had been reporting that as one of my pages winning that query. Search Console has a separate AI features report, which gives citation counts per URL. Different provider ... ... that is flat at a handful a week, there is no traffic to protect on either URL, so the redirect is cheap to try and easy to undo if that line ever moves. Comment from Alex Goldwyn(@alex_goldwyn): Mine

... 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 ... ... plain text with something like BeautifulSoup, you delete that evidence and flatten a concealed instruction into ordinary prose. You are not cleaning the input. You are laundering it. Almost every ingestion pipeline I have seen, including one of mine

... Mine ranked nowhere on launch day. Barely any upvotes, quiet comments, felt like a non-event. A month later: 1600+ users, 500+ signups, 250+ people who came back without being asked to. Zero ad spend, none of it came from PH. My honest take, this platform is mostly tech people, developers, builders, saas folks. If your product's for that audience, a strong launch here can genuinely be everything. If it's not, like mine

... about what those pages are allowed to load: no third party hosts, no analytics before consent, a content security policy, whatever yours is. Now look at what that rule actually iterates. Not what it says. What it loops over. Mine is a test that forbids any public page from blocking its own render on a host I do not own. It reads its page list out of the application rather than from a hand picked array, which was already ... ... page does a stranger reach on your site that your own definition of public does not include, and does your rule iterate a list somebody wrote, or the route table itself? I have not found a way to make mine

... party in the conversation with nothing to sell you, and that it's the version I'd have to defend if I were wrong. The second document is whichever GEO study is in your inbox this month. Mine was Previsible's 2026 State of AI Discovery - 6.77 million AI-driven sessions across 166 sites, November 2024 to May 2026, and genuinely the longest-running dataset of its kind. Its findings are plausible: Google's own AI surfaces drive more ... ... 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

... 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

... over twenty-five years of it on TV productions and it's what I am first, not a side note to something else. Photography came later, on my own time, and it's stayed exactly that: mine, separate, the thing I do because I want to, not because someone's paying me to. Rome to the set I studied at the Institute of Cinematography and Television R. Rossellini in Rome, and started working as an assistant cameraman ... ... post it. A finished shot, going nowhere, because the words never came as fast as the image did. The gap wasn't talent and it wasn't effort. It was that writing and shooting are different muscles, and mine

... built a benchmark that runs nine WordPress backup plugins head to head on identical sites. One of the nine is mine, and it currently sits second. That is an obviously compromised position, so I have spent more time on what makes a vendor-run benchmark believable than on the measuring. Three things I landed on, and I am not sure any of them are enough. Publish the runs you threw away. 581 completed runs, 180 scored, 401 excluded, and every ... ... reason attached. Most are boring reruns rather than embarrassing failures, and saying so is part of the point: 360 superseded, 41 excluded for cause. The honest number includes the boring ones. Say when a lead is not a lead. Mine

... newest one or two means people.Cross it against a surface only humans reach. My repo had 81 unique visitors while the registry was reporting 2,411 downloads. Thirty to one.Look for a spike with no correlated movement anywhere else. Mine had 1,542 downloads in a single day and the visitor graph for that day is flat. Nothing happened. A mirror woke up. Then work out what you cannot see at all, which was the uncomfortable one for me. Anyone ... ... unique cloners, 116, and CDN file requests, 95 a week. Both far smaller than 2,411 and both far more likely to be a person.I still show a number on the site, because everyone does and a visitor reads mine

... Founders ! drop your startup below :) I'm gonna check every single one and give you feedback. Here is mine : Murmell , it's basically google docs for AI coding agents, you can collaborate with your friend in the same board and use multi providers coding agents in the same rooms, and all that, in the cloud. So you can close your laptop and your agents will be still running :) link : Murmell Comment from Moss Ab Mirande-Ney(@moss_ab_mirande ...

... watermark in text. So does provenance matter for code, or only for content? Where I've got to so far, and I'd change my mind easily on this: For what I ship to users, the output is mine regardless of what typed it, so a label would be noise. For anything someone else has to maintain, I'd want to be told, because generated code tends to fail in a different place: the structure is plausible and the edge ... ... cases are wrong. For client work, it's their question to ask rather than mine

... threads are now outscoring every comment I have ever left combined. Still missing the one line almost everyone ranked above me on the leaderboard carries. Verified Account, plus 20, one time, no writing involved. Checked a dozen profiles above mine and all of them have it. I do not, and claiming it would be the single biggest jump available to me right now, bigger than my next several threads combined. If you have claimed it, what did the actual flow ... ... answer the flow question from the unclaimed side, because I have been through it twice and still do not have the line. The flow itself: form on the verification page, work email, GitHub, LinkedIn. All three of mine

... same answer independently oldest stays. I wrote the test, then broke the rule on purpose to check the test would notice. It reported zero survivors. The tally: forty lines of guard, nine corrections to get right, four of them mine and introduced while fixing the previous one. Free, universal, macOS 15.6+: https://netfox.app What's the most misleading bug report you've ever written for yourself? I'll go first mine

... thing I ve been thinking about while building is that our real competitor often isn t another startup. It s the spreadsheet someone already has, the manual process that s good enough, the employee who s always handled it, a messy collection of tools, or increasingly, just opening ChatGPT and asking it to figure something out. So I thought it could be fun to turn that into a little product arena. Put your product in the arena: [Your product] [what ... ... free. Maybe people don t realize they have the problem. Or maybe your opponent is genuinely the better option for some users. I think those answers are a lot more interesting than comparing feature lists between startups. Mine

... people in one Reddit thread called a comment of mine AI generated, in the same register I have used across every platform for weeks without issue. Went back and audited the account instead of arguing about it. The pattern that got flagged was not length or structure. It was confident mechanism reasoning with no first-hand marker in it, nothing that said I actually did this. Comment karma on that account went from 1 to 26 over two weeks once ... ... every comment started carrying a real number or a real mistake instead of just an argument. Curious whether anyone else here has found a similar tell for what reads as generated versus

... finally read the whole study instead of the headline, and the actually useful findings are different from what got shared. First, in their realistic-config lane (instruction file + several MCP servers), the "light" harness came out HEAVIER: ~90.8k vs ~75k. A 72KB instruction file alone added ~20k tokens to every request, on both harnesses. Their own conclusion: configuration, not the harness, accounts for most of the production bill. The harness sets the floor, you set the ceiling. Second ... ... close to a fixed cost. An unstable small one can out-spend it. Size isn't the sin, churn is. Third, session shape flips the winner anyway. On a multi-step task the heavy harness finished cheaper (121k vs