... suspect it's tiny, which would make the cap mostly peace of mind. Comment from Mona Truong(@monatruong_murror): @paul_kovalenko   Cost per completed task is a much better unit, agreed, and the two-calls-versus-twenty spread is exactly why per-call averages mislead. On your question, I can't give you a clean before-and-after on usage around a limit, and I'd rather not guess at one. Your rationing point is the one that worries ... ... than around compute. Paul's cost per completed task is the right framing for us too, since a finished campaign is the only unit that makes sense to measure. (I work on BlueVeta.com.) Comment from HongSJ(@cubig_sj): The axis

... suggestions. The differentiation always turns out to live one layer down, in which single signal the product picked to be good at and that's also where the category quietly fails. Look at what's actually out there. One competitor nails color harmony genuinely good color theory, matches pieces that visually work together. Another does a chat-style interview to build a style profile. A third tracks wear history and layers in weather data so you stop repeating the same ... ... five outfits. A fourth proved the whole concept works, but shipped iOS-only. Each of these is a real, defensible product. None of them constrain on more than one axis

... moved a folder. 0.34.0 is four changes that are the same sentence from different angles: a security signal is only worth reading if it fires on what happened, not on what you did to your own repo. An axis that cries wolf is one you learn to skip, and then it's worth nothing on the day it's right. The workspace fix: the lockfile parser keyed a package by its path, so a monorepo reorganisation produced a screenful ... ... pleased about is the one nobody would have seen: changing that key would have made the first scan after the release report every install-script workspace as newly arrived. The upgrade would have been indistinguishable from the thing the axis

... gradient shine sweep that loops infinitely without impacting CPU threads. We engineered this button to recreate the satisfying physical feedback of pressing a mechanical keyboard switch. Using nested CSS box-shadow layers and a clever translation on the Y-axis via Tailwind's translate-y classes, the button achieves a 3D isometric look without WebGL. The "shine" effect is powered by an optimized CSS linear-gradient animation that runs entirely on the GPU, ensuring that even on lower-end devices ...

... trigger. A workflow reacting to discussions on a repository with discussions disabled is not an exposure. It moves scores both ways fewer findings where a setting makes a trigger unreachable, more on private and internal repositories, where the axis used to go quiet and where a read-access user is exactly the supplier worth worrying about. Two tips for getting the most out of it: pipe the activity feed through jq and you have a personal work ...

... does is a finding. The same version shipping different bytes is a heavier one no upgrade explains that. An ordinary version bump gets listed and scores nothing. Tip : it rides the same baseline as the rest of the delta axis, so the 30-day window and the "compared against a baseline recorded N days ago" line apply here too. Hit s on any repo in the Repos tab. Second tip: if your lockfile is pnpm's or Yarn ... ... files churn constantly; the install-script subset moved twice, and the same version was republished zero times. That zero is what lets the sharp case carry real weight instead of teaching you to skip the axis

... introducing a kind of entropy in your codebase, duplication, mismatching docs, too big functions, etc... so I often run an "entropy-reduction" loop, where it scans the codebase architecture, the performance, security, doc drift, etc.. a few more axis, which then afterwards I use n subagents to create plans to address those points, this helps a lot with keeping things under control. Comment from Kabir Suleman kusfa(@kabir_suleman_kusfa): I usually trim everything down to the few errors ...

... same action can have very different risk depending on context. Refund is one verb, but $5 to a known customer and $50K to a new destination are obviously not the same decision. Same with sending a message internally versus sending it publicly in someone s name. Do you add conditions around the verb as well, like amount, destination, user, environment, or some other context before the agent is allowed to run it? Comment from Eric Savoie(@eric_savoie): Mostly approval ... ... payee is the actual risk, not the number. Same logic on messaging, an internal note runs free, anything that goes out publicly in my name gets a gate even if the content is trivial. So the verb decides which axis

... seconds. the other one spends a first impression, permanently, on a business we do not get to approach a second time. i had been gating on whether a thing could be undone. that turns out to be the wrong axis entirely. it fires constantly on things nobody cares about, like writing a log line, and it stays quiet on the ones that actually hurt. so the gate moved onto permanence instead. anything reversible inside a minute runs unattended. anything that ...

... Every fix we've seen sorts into two buckets. Compressing acceptance lighter or no review converts review cost into incident cost, payable later with interest. Leveraging acceptance is the unglamorous bucket that works: smaller PRs (size is the axis authors control, and it moved +51% the wrong way), artifacts that arrive carrying their own evidence test output, the exact command to re-run so review shifts from reconstructing intent to checking claims, and machine passes before human attention. One honest ...

... much! I inserted a form with the feedback. :) Comment from Sabber Ahamed(@sabber_ahamed): @busmark_w_nika   @abdullah_javaid3  the random-id-on-activation trick is the right call. weekly active hides exactly the churn-vs-loyal split you're describing, twenty once-and-done and twenty daily users report the same number. Comment from Myrto Papagiannakou(@myrto_p): @david_martin_suarez  I would love to hear more about that. We might be able to help. Comment ... ... audit checks the interface against known patterns. It won't tell you whether the tool solves the problem for the person it's actually for. Comment from Rabnoor Singh(@rabnoor_s): The number is the wrong axis

... Googlebot at all" and only needs existence. It would quietly ruin any question shaped as "how many" or "did this specific request get through". Comment from Omri Ben-Shoham(@omri_ben_shoham1): Good writeup - the distinction between blocking Training vs blocking Agent is the part people are going to get wrong for a while. One thing worth adding from the operating side of an agent, not the site-owner side - when an agent gets blocked, it usually doesn't fail ... ... answer down somewhere the next person finds, per site, because the thing that made this bite is that it was a decision nobody recorded making. Comment from Mona Truong(@monatruong_murror): @serdarships  Varying the client is the axis

... question back at you in synonyms is bad whether a person or a model produced it. A five line reply that answers something nobody else answered is good regardless of what wrote the first draft. Flagging the wrong axis just teaches people to sand off the tells rather than add the substance, which is the opposite of what this is meant to encourage. Comment from Arnold Oshenye(@oshylabs): @chrismessina   @rabnoor_s  that api detail explains a lot actually ... ... long as it isn t over reactive, I think it could be good! 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

... think we have different usecases :) Comment from Nika(@busmark_w_nika): I was more used to Claude Code. Codex is okay, but not good for chatting the way I liked :D Comment from Rabnoor Singh(@rabnoor_s): the axis nobody picks in these threads is which one you can constrain, and for me that decides it more than raw model quality does. what i care about is whether i can write down, in a file the thing reads before ...

... completely normal it fires on a tag push, so the only people who can reach it are people who could already push tags. if that scored, a big chunk of github lights up and everyone learns to ignore the axis inside a week. so the rule became: capability isn't a finding, capability reachable from untrusted input is. same file triggered by pull_request_target now it's a privilege-escalation gadget. and a self-hosted runner only counts ... ... self-hosted, not just because both happen to exist in the same repo otherwise the report claims outsider code can run on your hardware when it can't. on top of that there's a hard ceiling: 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 Alper Tayfur ... ... worth more than the compliment was. Comment from Venkat A(@venkat_a2): Draft by default is what we do, for what it's worth. Nothing we generate publishes itself, a person has to press the button, so on your axis

... gated on the caller being reachable from a fork trigger. The reasoning was noise control a repo calling a third-party reusable workflow on a tag push is ordinary, and listing every one of those is how an axis gets ignored. Read back through your sentence, though, every one of those unlisted calls renders identically to a chain that terminated safely. Which is the collapse, on the exact axis

... write the comment, or is it mostly trial and error for you too? Comment from Abdullah Javaid(@abdullah_javaid3): @lily_foster2  That's basically the tradeoff I was trying to name and failed to. Fresh and quiet versus busy and buried, and I picked on gut feel every time, not a rule. Comment from Nika(@busmark_w_nika): @abdullah_javaid3  But how did you know how many impressions the comment got? Only threads show the impressions, not comments ... ... replies were addressed to you rather than to the OP. If it is zero, that was the worst of the four, not the best. Comment from Abdullah Javaid(@abdullah_javaid3): @busmark_w_nika   @pamela_arienti  The generic vs

... 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 ...

... flow look like? Trying to work out if it is friction in the process or people just not knowing the line exists. 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 ...

... Genuine question, since Writeara lives exactly on this line: do you treat every destination as one publish-gate, or does the gate's weight scale with the destination's reach, a push to a personal blog with 12 readers versus a client's site with 100k? Because "reach, not reversibility" implies the gate should be proportional , and I can't tell if that's worth the complexity or if a single publish-boundary gate is the right 80/20. Curious where ... ... because it's dumb, but because the radius is determined by capabilities it isn't even looking at. That's the strongest version of the "worst judge of its own reach" point I've seen. And your authority-armed-versus

... quietly break a promise. Comment from Raffay Sajjad(@raffay_sajjad): Building a finance app, so I think about this constantly. My take lines up with Rabnoor's, not sensitivity, whose it is, but I'd add a second axis: what can you reconstruct from it over time. A device model leaking is nothing. Transaction categories over six months is basically a biography of someone's life, when they're broke, what they're hiding from a partner, if they ... ... policy says 'X is collected,' there's zero technical barrier stopping them from harvesting Y behind the scenes. 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

... 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 Raj Nagulapalle(@rnagulapalle): thank you @galdayan ...