... 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 ... ... last eighteen months assumed. Who paid for this, and why I'm citing it anyway. Two of the three authors previously held postdoctoral positions at Microsoft and now work as paid research consultants for the company. That's disclosed

... neither is collapsed into a verdict. What I am genuinely unsure about is whether any of that is enough, or whether a benchmark run by someone selling one of the entries is simply not credible no matter what it discloses, and the only real answer is handing the method to someone with no stake in it. So: if you have run or read one of these, what would you need to see before you believed a number published ...

... damon__mercado): Maybe the real issue isn't whether AI made it but whether someone checked if afterward. That seems true for both writing and code. Do you think human review is enough or should AI use also be disclosed? Comment from Elen Tovmasian(@elen_tovmasian): I think the difference comes down to expectations. When we read a comment, we expect a genuine human perspective. When we use a product, we expect it to work reliably and securely, regardless ...

... keep coming back to the same question with AI phone and operations agents: what should be disclosed before the agent is allowed to do anything that affects another person or system? I do work in this area, so I am biased toward over-communicating. My current instinct is that the minimum should be explicit identity, whether the conversation is being recorded or summarized, what actions the agent is allowed to take, when a human can step in, and some kind ... ... reversible low-risk tasks, but anything that creates a financial commitment, changes identity details, or sends a message the user would not want quoted later should force explicit confirmation. Then the receipt becomes the audit layer: what the agent disclosed

... standard editing. If you deploy emotion recognition or biometric categorisation, you have to tell the people exposed to it. Notably you don't have to explain the purpose just the operation. If you deploy a system generating deepfakes, you disclose. Same for AI-generated text published to inform the public on matters of public interest unless it went through real human review with someone holding editorial responsibility. Spell-check doesn't count. All of it has to be clear, distinguishable ...

... even less comments on launches ? Crickets amongst the AI bots? A genuinely good use case is using AI to improve non-English-speaker's contributions. In those cases, it would be lovely if the commenter would self-disclose so that the AI label isn't held against them. I wonder what they're using behind the scenes for this like @Pangram or something else? Comment from fmerian(@fmerian): Love it. The Product Hunt community isn't just the most curious ... ... very logical way to restrict at least part of it! Comment from Ryan Miller(@ryanwrites): @alieksia  I'm worried about false positives too. I would prefer to be able to report something as being AI vs

... fuzzy on purpose, and fuzzy-on-purpose is exactly where indie products live. If you're in scope, here's what SB 243 actually requires, and I want to make the case that none of it is unreasonable. One: disclose that it's AI when a reasonable person might think they're talking to a human clear and conspicuous, not buried in a terms-of-service page. Two: have a real protocol for self-harm and suicidal ideation ...

... users , because the coverage of this was all about the parts that got delayed . Article 50 of the EU AI Act applies as of 2 August 2026 . In plain terms: An AI system that interacts with people has to disclose that it's AI (unless it's obvious from context) that's your support chatbot, your onboarding assistant. Synthetic media images, audio, video has to be labelled as artificially generated. AI-generated text published to inform the public needs disclosure ...

... actually failed, spanning 22+ countries and industries from fintech to biotech to retail. Each entry includes funding history, investors, a sourced explanation of what actually went wrong, a timeline, and a plain-language lesson, built to be a genuine alternative to expensive tools like Crunchbase that treat failure as a footnote rather than the actual subject. Every entry is manually researched and verified against public reporting, not scraped, and where a figure was never publicly disclosed

... shipped work where 90%+ of the code came from prompting an agent, with me mostly reviewing, testing, and directing the architecture. Never explicitly disclosed that to the client, they just know it got done fast, works well, and stayed in budget. Is that dishonest, or is it no different from using a framework, a library, or a junior dev and not itemizing every line's origin? Nobody used to ask what percentage of a website a developer typed themselves versus

... agent didn't just act, it acted with someone's credentials, someone's access, someone's blast radius. and when it goes wrong, the audit trail usually can't tell you what the agent was actually authorised to do versus what it happened to have access to. those are different things and most systems conflate them. honestly that's a big part of why we're building MonoCloud the way we are. scoped identity for agents so the blast radius ... ... there's no shared webhooks folder they both sit under. so the glob fix doesn't even apply to the three i have today, they were never colocated in the first place. which proves your enumeration vs

... user and any friends they add. The free app, which offers some paid features, has about 4.3 million monthly active users, according to Sensor Tower data. The previously unreported acquisition closed this spring, Midjourney said. It declined to disclose financial terms. This is big: CEO @banu_guler is now Chief Design Officer at Midjourney. Comment from Varun S.(@varuns): Wonder what the angle is here. Looking forward to the shifts. Comment from Alpaysh(@alpaysh): incredible! Comment from Oluwole Zacheus(@emmadagold ...

... moved and tmux reattached elsewhere are the same failure underneath: the identity you resolved against stopped being stable, and nothing told you. Which is the argument I have been making all week about session memory generally, so I should disclose the obvious. I hunted Atuin on here today, which is the shell history side of exactly this: history that keeps the directory, the exit code and the duration instead of a flat file that forgets which session it belonged ...

... animation and 8 seconds of silence. Our engineer thought it was a bug when he first saw the spec. "You want me to add... a delay?" Yes. On purpose. Here's why. We ran an experiment: instant AI response vs. an 8-second pause before the same response. The content was identical. The only difference was time. Users who got the pause rated the AI's reflections 40% more helpful. Same words. Same insights. They just felt more considered ... ... honestly surprised us too -- we expected maybe 10-15%. It made us rethink a lot of our other interaction patterns. Comment from Mona Truong(@monatruong_murror): @alieksia   Really sharp question. We haven't run a controlled test on disclosed

... like selling myself either, so I don't. What I do is write comparison pages between the competing products and mine, and put everything into ranking those. The people who land on them are already interested in a competitor, or already using one and unhappy with it. They read what each product does and what it costs, and if mine comes out as the reasonable choice, the conversion happens sooner or later. If you don't have much ... ... think "salesy" isn't really about word choice it's about whether you're answering something someone asked, or interrupting something they weren't asking about. Curious if others have found the line shifts once you have actual users vs

... play out there? Comment from Mira Charkawi(@mira_charkawi): I'm currently using phantombuster, gojiberry & Clay. Comment from Moeez(@moeez1): @sudipta_biswas4  Fair question. From what we ve seen, people don t really react to AI vs human , they react to whether the call is relevant and well-timed. And yes, if someone asks, it s always disclosed

... questions likes that under 10 seconds, end to end, presenting you answer, charts (if needed), all data used and a non-technical explanation what is what (if you drill down to it). When we do it, we don't disclose your data to LLM, while at the same time still using excellent coding agents in realtime to build, run and show you a one-off application to analyze and answer your question, all in under 10 seconds! Under the hood ...

... funding, I had to automate everything, therefore I relied on a cross environment team of AI assistants, that I, painfully and traumatically, intuitively guided to build deterministically. Since patent is pending, I cannot reveal too much, but I have disclosed enough for technical people to understand the architecture. You can also feed my announcement landing page URL to an AI assistant, so it could explain the technology ...

... have actually diverged in outcome in the ledger so far and if it's inferred, what's your cold-start before a single divergence has been observed? Comment from Alexander Bering(@alexanderbering): @aadilghani  The reason "edit vs. tell-it-to-revise vs

... economics just don't math. A few things I'm chewing on: Is usage-based or credit-based pricing now inevitable for any AI product with an agent layer? Anthropic claims Max 20x gives 20x the usage, but the disclosed numbers work out closer to 6x on Claude Code hours. Are we entering an era where "20x" means tokens, not outcomes? Cursor got dragged for communicating their pricing change poorly. Anthropic seems to be getting ahead of it with ...

... without acknowledging it beforehand is quite disrespectful in my opinion. In addition to trust, decency that comes with being forthright was lost. Definitely not the best way to establish a new relationship. Now, I wait to see if they disclose the recording before I chime in to ask about it at every video call. I think it is a form of etiquette to get consent. Comment from Rohan Chaubey(@rohanrecommends): @byalexai  Thanks for the thoughtful reply. I am using ... ... handle information. Comment from Farhad Asbaghipour(@farhad_asbaghipour): I d probably drop immediately trust is gone at that point. It s interesting how many modern workflows are becoming record first, ask later. Curious how others balance efficiency vs

... answer first. If they find it immediately, they cite it. The structure makes extraction easy. Step 2: The Comparison Page Template Comparison pages drove 40% of our citation growth. Here's our exact template: Title: [Brand X] vs [Brand Y]: Which is better for [use case]? Quick verdict (100 words) Who wins and why One sentence summary Feature comparison table Feature Brand X Brand Y Feature A Feature B Feature C Price XX YY Detailed breakdown (by feature) How each ... ... margin of error. Result: This block was cited verbatim by Perplexity within two weeks. Step 5: The 90-Day Update Cycle We analyzed citation decay across 500 pages. The drop-off is steep: Months Since Update Citations (vs

... think we all lie every day (even when we say "Good night" to someone, and we don't even have to mean it sincerely). We also lie on our resumes , we don't fully disclose everything on our tax returns , and we sometimes fake income . And that's what I want to talk about... Cluely CEO Roy Lee admitted that the $7M ARR figure he gave TechCrunch last year was fabricated , calling it the only thing he's "blatantly" lied ...

... social media, and manages my spreadsheets while i sleep. sounds insane but after a few weeks you stop thinking about it the same way you stopped thinking about autofill having your credit card. the real question isnt trust vs no trust, its what guardrails you set up. mine cant send emails without me approving the draft first. cant post publicly without a review step. but it can read, organize, research, and draft freely. thats the line that works ... ... smart" to be useful. Comment from Navanita Devi(@navanita_devi): Dedy's point is interesting, the opacity problem is real. Most people conflate "AI having access" with "AI having unchecked control." They're very different things. Read-only access vs

... loved one s voice using recordings that were fully consented to in advance. It is not about replacing someone or simulating the dead. It is about giving people a voice when silence becomes unbearable. Every call is disclosed. Every voice is permission-based. Every message is controlled by the person receiving it. Some people use it to say goodbye. Some to feel steady on hard days. Some just to remember what it felt like to be known ...