... that attack class from Ionut-Cristian Florescu's writeups he's the maintainer whose own repos got hijacked rather than a filename list, which is why renamed variants still trip it. but i should be straight about the delta axis specifically: it can't be validated that way by construction. you'd need a fingerprint from before the compromise, and a cleaned repo only ever gives you the "after". best i have is synthetic doctoring a baseline and confirming ... ... lesson, and the signature thing is exactly it. What they cant teach is false negatives, which in a security tool is the expensive direction to be wrong in. Best example came from neither: a unit test asserting "no single axis

... here's what we do, here's where we're unreliable, here's the thing we put in front of you before it goes out. That's a harder page to write and a much harder one for a competitor to copy, because it means you actually went and found your failure modes. Curious whether anyone has tried this and what it did to conversion. My guess is it costs you signups and buys you retention. Comment from Venkat ... ... entirely and nothing in your analytics will ever surface it. Which probably means it only ever gets fixed by someone deciding to, not by measuring it. Comment from Asad M.(@asadmalik901): @ark_y_k  Reversibility is a better axis

... show up when you search our brand name. A comment is a page in Google's index long after the views stop moving. Your 9-view thread and your 221-view thread may age very differently on that axis. For context, I run sourcing checks from Yiwu, so PH threads are not where my buyers sit anyway. Which forces exactly the discipline you describe, researching the thread before writing a word. Comment from Adekemi A.(@_adekemi_a): i've stopped ... ... days, then read the number PH shows on the thread itself. No way to isolate just the comment's own reach, only the container it sits in. Comment from Abdullah Javaid(@abdullah_javaid3): @jernej_jan_kocica  The read versus

... this be undone" to "how far does this reach", which is a real improvement. Irreversibility over-fires on writing a log line and under-fires on a migration you can only roll back with downtime. Reach is the better axis. But every gate in that conversation, and every gate I have seen shipped, evaluates a single action in isolation. Sending one email to one person is low reach. It passes. Four hundred of them inside twenty minutes is a mass ...

... never the lever I hoped it was, and I found that out from your breakdown rather than another rejected test post. Comment from Jernej Jan Ko ica(@jernej_jan_kocica): @abdullah_javaid3 Fair, cost per point is the better axis and I was comparing sizes. The distinction I would keep is that it is one-time. Cheap and finite means do it now and stop thinking about it, but that makes it a floor, not a lever. Once it is claimed ...

... only afterwards, and it is a property of your organisation rather than of the call. So it cannot be an input to an approval decision at all. It can only change how much you let run between checks. One axis is a gate, the other is a leash length, and treating the second like the first is how it stays underbuilt. On the org boundary you have the cleaner version of something I have been circling in another branch here ... ... Nothing in the signature knows who asked, so "not on that authority" isn't expressible. Today that's hidden rather than solved: Eli only acts when someone asks in the sidebar, so the trigger is always a person. The axis

... very grateful for you guys giving valuable advice and insights. Comment from Willis Liao(@willis_liao1): @rabnoor_s  Well said Comment from Willis Liao(@willis_liao1): @santhosh_kumar09  True, and because comparisons matter, although local alternatives might still collect your data when connected to wifi, the risk and the data stolen is way smaller compared to cloud apps. Comment from Willis Liao(@willis_liao1): @ark_y_k   True, the user usually doesn't really know what they ... ... what they should be concerned about, at least not until something unfortunate has happened to them or other people. Comment from Willis Liao(@willis_liao1): @irahimiam  True, that's why more and more people are choosing local alternatives

... have to move from opinion-based review to evidence-based verification. This means building guardrails (what I call the claim checker) that bind the original intent to deterministic evidence: ACs (intents) checks, the builds, the tests, the five-axis code review, the scans, the blast radius of the code change, the docs coverage, requirements are complete and don't contradict each other, and the monitoring coverage in the code. AI can advise. But only evidence should decide whether to merge ...

... inside your app if they can only feel understood while they're talking to your model you haven't solved loneliness. You've made it more comfortable to stay lonely. You've built a very sophisticated waiting room. The alternative is to design for the exit. Concretely, that means three things. Point users back at real people, not deeper into the app. The most valuable thing an emotional AI can do isn't be a better listener than your friends ... ... from Mustafa Arian(@thenameisarian): the design for the exit framing lands. the missing piece for teams trying to implement it is the measurement problem gal pointed at. self-reported outcomes are weak signal. in-app metrics are the wrong axis

... asking for contents: write whose caller passes nothing holds nothing, and doesn't score. permissions: {} in the caller is the sharp version: grants nothing, but it is a declaration, so the repository default doesn't apply either. The axis also has a hard ceiling below the "suspicious" threshold, so capability can't move a verdict on its own however many findings a chain produces that constraint is the only reason I was comfortable letting it see more. Cross-repo calls ...

... back-to-back redelivery, the retry storm, which catches the race and the naive dedupe but not your TTL case. fast-forwarding past a dedupe window means controlling delivery time, not just order, and that's the axis i'm building toward next, precisely because the window is where the bug hides once you've fixed the obvious one. so honest no today, and it's exactly what i want it doing. Comment from Arnold Oshenye(@oshylabs): The part that ...

... demand question, and favour testers cannot answer it at twenty or at two hundred. Of the people testing now, how many arrived on their own rather than being asked? Comment from Arnold Oshenye(@oshylabs): Bharat's point about views versus responses is the one I would add to. I launched a niche legal tool for UK renters and the number that actually told me something was not signups, it was how many people who read the free analysis came back ... ... release. That's awesome, keep it up And I hope you get so many users that it becomes a "new problem" for you in a good way Comment from Rabnoor Singh(@rabnoor_s): The number is the wrong axis

... CLAUDE.md file grow too long. Comment from Omri Ben-Shoham(@omri_ben_shoham1): @Gal Dayan the blast radius framing is better than mine honestly, I was thinking about it purely in terms of "how ugly" which is the wrong axis entirely. a messy pure function is basically free to clean up because a bad refactor there just fails its own tests loudly. the scary case is exactly what you said, shared state or external calls where the tests all stay ...

... Same queue, totally different urgency. Telling those two apart automatically is the piece I don't have clean yet. In your setup, how are you deciding which of the 20 percent is allowed to wait for a window versus which should break into your focus? Comment from Aadil Ghani(@aadilghani): @virko_kask   Agent-state instead of task-type as the urgency signal is better than what I had, and I think you just moved my whole priority model. Category ... ... decides if it may reach me, idle-vs

... features, way more surface area for that to happen unnoticed. Comment from King Joshua Marcos(@moddy): Honestly just trial and error early on. Let it fully own something low stakes, watched what broke, adjusted from there. The billing/auth vs UI split came out of getting burned once on something billing-adjacent, not from planning it ahead of time Comment from King Joshua Marcos(@moddy): @donna_gerrard  Would probably help me a lot, I don't have one written down ... ... compiles fine and looks totally reasonable while being numerically wrong, and I won't catch that without actually running the eval. Comment from Sabber Ahamed(@sabber_ahamed): billing/auth strict, everything else looser is basically my rule too. the axis

... license A quality B maintenance Tool-definition quality: A, 4.4/5 tool descriptions averaging 4.5/5 across all 39 tools What made it actually useful, not just a badge: the first test came back 3.9 with a per-axis breakdown showing exactly which tools were weak. We rewrote every tool description against their rubric, republished, and the same-day re-test came back 4.1. A neutral grader + a fast fix loop beat weeks of my own polishing ...

... server URL and connect Basedash to AI tools like ChatGPT, Claude, and Cursor and made the URL easy to select and copy manually. Added Claude Sonnet 5 and Claude Fable 5 to the model picker. Rotated crowded x-axis labels on vertical bar charts so they stay readable instead of overlapping, with extra room reserved so nothing gets clipped. Refreshed chat attachment previews with rounded image thumbnails, richer file cards, consistent sizing, and a hover-to-remove control on every ...

... first one, but my instinct says the second ages better because nobody has to hold the gnarly part in their head six months from now. curious how you'd call that one, since it's not really the same axis as "fewer moving parts." Comment from Gal Dayan(@galdayan): @bickov  "failures got quieter" is the whole thing in three words. that's the actual cost nobody prices in when they measure agent speed against human speed, a human ... ... version against that bar. Selection after the fact is reviewing, specification before is designing, and the second is where you actually control quality. When I do still want variants I make it deliberate, two that differ on one axis

... upvotes. Comment from Hana Salazars(@hana_salazars): I would want a simple explanation of every permission and why each one exists. That would reduce hesitation before installing. Comment from Anuj(@getosmo): Most answers here are attacking the evidence axis, so let me push on the other half of your question - what would still stop me enabling it. I build a consumer product where people forward their actual phone calls to it (OsmO), so the ask is arguably more invasive than ...

... leaves the device Comment from Haider(@lilhadi): @ceciliatran  it's quick for most people because you see what it understood before you hit add, so there's nothing to trust blindly Comment from Haider(@lilhadi): @galdayan  the axis is OS more than typing style on macOS 26+ the on-device model catches almost all of the tail, so cloud rarely fires no matter how messy you type pre-26 (which are older macs) has no on-device model ... ... wrong one for something time-based. Comment from Gal Dayan(@galdayan): my graveyard entry is Things 3, not because it lacked features but because opening it felt like starting a mini project just to write "buy milk". the capture-vs

... universal format. Everyone already knows how to read one. It's a blank canvas. It doesn't assume what you're doing, so it never gets in your way. The newer tools usually beat it on one axis (collaboration, AI, visualization) but lose on flexibility or familiarity, so people keep one foot in the spreadsheet anyway. I see this constantly. Teams have a slick tool for the dashboard nobody opens and a messy .xlsx where the real work happens ...

... expanding. Manga and webtoons may come in future updates - each format deserves its own reading logic - but right now the priority is doing comics right. Reading that respects the medium (free) Full-screen reader with horizontal, vertical, and dual-axis modes Ambient mode - soft background color pulled from the page art Deep zoom, smooth page turns, pull to jump to the next issue without breaking flow Progress saved locally, covers and metadata enrichment Premium (optional - one lifetime unlock, no monthly ...

... from 3 brands whose use case I'd already validated. The signal compounds, every reply teaches me what the next 50 leads should look like. The piece that's hard isn't finding lookalikes. It's defining the "alike" axis correctly. Two brands with identical Shopify revenue, traffic, and ad spend can be completely different customers if their creative cycle is 1 week vs

... numbers toggle. By default Basedash keeps big values compact (like 1.2M), but when you need the exact figure you can flip a chart to show the full number in the primary value and data labels while keeping the axis labels compact and readable. It s an easy way to make number and KPI charts precise without cluttering the rest of the chart. Dashboards stay visible while their data refreshes Charts on dashboards now render their cached data immediately, even ...

... guess which element on a busy screen I actually meant, and it edits the wrong one. Then I'm re-prompting, which is its own iteration tax. Curious how much of your rework is this specific thing versus raw model quality. And what's your fix, crop hard, or describe the element in text? Comment from Alexander Bickov(@bickov): @alopez3006  That's the cleanest version of it I've heard, project context as a retrieved operational layer ... ... element thing, it's orchestration drift, sub-agents wandering off. One small overlap: when each sub-agent re-reads the screenshot, they can each interpret it a little differently, so a shared structured reference at least removes that one axis