[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-grok-build-turns-xai-into-coding-agent-en":3,"article-related-grok-build-turns-xai-into-coding-agent-en":36,"series-ai-agent-7776d8ed-2de2-4421-bcaa-d5b731a77a4c":89},{"id":4,"title":5,"content":6,"summary":7,"source":8,"source_url":9,"author":10,"image_url":11,"keywords":12,"language":18,"translated_content":10,"views":19,"is_premium":20,"created_at":21,"updated_at":21,"cover_image":11,"published_at":22,"rewrite_status":23,"rewrite_error":10,"rewritten_from_id":24,"slug":25,"category":26,"related_article_id":27,"status":28,"google_indexed_at":29,"x_posted_at":10,"tweet_text":10,"title_rewritten_at":10,"title_original":10,"key_takeaways":30,"topic_cluster_id":34,"embedding":35,"is_canonical_seed":20},"7776d8ed-2de2-4421-bcaa-d5b731a77a4c","Grok Build turns xAI into a coding agent","\u003Cp data-speakable=\"summary\">Grok Build is xAI’s first coding \u003Ca href=\"\u002Ftag\u002Fagent\">agent\u003C\u002Fa>, and I break down how to use the idea without the usual agent fluff.\u003C\u002Fp>\u003Cp>I’ve been watching \u003Ca href=\"\u002Ftag\u002Fai-coding-tools\">AI coding tools\u003C\u002Fa> get more ambitious for a while now, and honestly, most of them still feel like they were built by people who’ve never had to ship a real repo under pressure. They can autocomplete a function, sure. They can even patch a file and brag about it. But when I actually need help, I want something that understands the shape of the codebase, the constraints of the stack, and the fact that I do not want to babysit a chatbot through twelve “great idea!” replies before it gets to the point.\u003C\u002Fp>\u003Cp>That’s why xAI’s \u003Ca href=\"https:\u002F\u002Fwww.pcmag.com\u002Fnews\u002Felon-musks-xai-launches-grok-build-its-first-ai-coding-agent\">Grok Build\u003C\u002Fa> caught my eye. PCMag’s write-up is short, but the signal is obvious: xAI is trying to move Grok from “chat model” into “coding agent” territory. I’ve seen enough half-baked agent demos to know the difference between a cool screenshot and a tool I’d actually wire into my workflow. This one matters because it’s not just about writing code. It’s about deciding, acting, and fitting into the boring parts of software work that usually get ignored in \u003Ca href=\"\u002Fnews\u002Fwhy-rust-community-threads-still-matter-en\">launch posts\u003C\u002Fa>.\u003C\u002Fp>\u003Cp>So I’m going to break down what this launch actually implies for developers, where the agent pitch tends to fall apart, and how I’d think about adopting it without turning my editor brain off.\u003C\u002Fp>\u003Ch2>What xAI is really saying here: Grok is moving from talker to worker\u003C\u002Fh2>\u003Cblockquote>“AI coding agents are on the rise, and xAI is the latest to join the race with Grok Build.”\u003C\u002Fblockquote>\u003Cp>What this actually means is simple: xAI is no longer positioning Grok as just a conversational model. It’s trying to make Grok do work inside a development workflow. That sounds subtle, but it’s a big shift. A chatbot answers. An agent acts. If the product is serious, I should be able to ask it to inspect code, propose changes, maybe even carry a task across multiple files without me re-explaining the whole project every ten seconds.\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779174862471-tn1e.png\" alt=\"Grok Build turns xAI into a coding agent\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>I’ve been burned by tools that look agentic but collapse the second the task isn’t toy-sized. They can rename a variable. They get weird when the job is “trace this bug through three layers of abstraction and keep the public \u003Ca href=\"\u002Ftag\u002Fapi\">API\u003C\u002Fa> stable.” That’s the line I care about. If Grok Build is real, the question is whether xAI is shipping a tool that can survive that line.\u003C\u002Fp>\u003Cp>How to apply it: treat every “\u003Ca href=\"\u002Ftag\u002Fai-coding\">AI coding\u003C\u002Fa> agent” announcement as a workflow claim, not a model claim. Ask what it can do across files, what context it keeps, how it handles diffs, and whether it can work with your repo instead of just talking about it. If the answer is fuzzy, the launch is mostly branding.\u003C\u002Fp>\u003Ch2>Agentic coding is useful only when it stops acting like a yes-man\u003C\u002Fh2>\u003Cp>One thing I keep seeing in AI assistants is this annoying habit of agreeing too fast. You suggest a direction, it says yes. You change your mind, it says yes again. That’s fine for brainstorming, but it’s useless when I need a system that can push back on a bad assumption. A coding agent should notice when my plan is brittle, expensive, or just plain dumb.\u003C\u002Fp>\u003Cp>That’s the real test for Grok Build, even if PCMag’s note doesn’t spell out every feature. Does it just generate code, or does it critique the task? Can it tell me that my approach will create a mess in the test suite? Can it suggest a smaller patch instead of a heroic rewrite? If not, I’m not getting an agent. I’m getting a verbose autocomplete with a startup demo.\u003C\u002Fp>\u003Cp>I ran into this exact problem when I tried using an assistant to clean up a legacy service. It kept producing “reasonable” refactors that were technically correct and operationally awful. It split one file into five, introduced naming drift, and made the bug harder to reproduce. The model wasn’t wrong. It was just unhelpful. That’s the difference between code generation and code judgment.\u003C\u002Fp>\u003Cul>\u003Cli>Good agent behavior: asks for scope, checks assumptions, and narrows the task.\u003C\u002Fli>\u003Cli>Bad agent behavior: confidently expands the task until you regret opening the tab.\u003C\u002Fli>\u003C\u002Ful>\u003Cp>How to apply it: when you test Grok Build or any coding agent, give it a task with a trap in it. Ask it to make a change that affects tests, docs, and one public interface. See whether it warns you about the blast radius. A useful agent should slow you down when it needs to.\u003C\u002Fp>\u003Ch2>“First coding agent” is a product milestone, not a finish line\u003C\u002Fh2>\u003Cp>PCMag frames Grok Build as xAI’s first AI coding agent, and that wording matters. “First” usually means the company is still figuring out what category it even belongs to. I don’t say that as a dig. I say it because first versions reveal priorities. Is the company optimizing for demos, or for daily use? Is it trying to impress developers, or reduce their friction?\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779174860347-ws31.png\" alt=\"Grok Build turns xAI into a coding agent\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>When a company enters the coding-agent space late, it has to answer annoying questions fast. How does it compare to \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Ffeatures\u002Fcopilot\">GitHub Copilot\u003C\u002Fa>, \u003Ca href=\"https:\u002F\u002Fwww.anthropic.com\u002Fnews\u002Fclaude-code\">Claude Code\u003C\u002Fa>, or the growing pile of agent tools people are already testing? What’s the boundary between model capability and product design? And maybe the hardest one: why should I move any of my work here instead of using the tools I already trust?\u003C\u002Fp>\u003Cp>I’ve seen teams get distracted by the “first” label and miss the real issue. The first release is rarely the one you keep. It’s the one that tells you whether the company understands developer pain at all.\u003C\u002Fp>\u003Cp>How to apply it: don’t evaluate the launch as a feature checklist. Evaluate it as a signal about product maturity. If the agent has weak task boundaries, poor repo awareness, or no sensible review loop, then it’s not ready for serious use no matter how polished the demo looks.\u003C\u002Fp>\u003Ch2>The only coding agent that matters is the one that fits your repo\u003C\u002Fh2>\u003Cp>Every coding agent pitch eventually runs into the same wall: real codebases are ugly. They have half-finished migrations, stale tests, weird naming, and one critical path nobody wants to touch because it was written by a contractor in 2021. A model can know syntax. That’s not the same as knowing your project.\u003C\u002Fp>\u003Cp>That’s why I care less about whether Grok Build can write a function and more about whether it can operate inside a repo with context. Can it read the right files? Can it follow local patterns? Can it avoid “helpfully” rewriting code in a style the team never uses? If it can’t do those things, it’s not helping me ship. It’s adding cleanup work.\u003C\u002Fp>\u003Cp>I’ve had agents confidently produce code that passed a casual glance and failed every important expectation. Wrong imports. Wrong assumptions about config. Tests that were “improved” into irrelevance. The annoying part is that none of that looks bad in a demo. It only shows up when you try to merge.\u003C\u002Fp>\u003Cul>\u003Cli>Repo-aware agent: respects existing structure, conventions, and test patterns.\u003C\u002Fli>\u003Cli>Repo-blind agent: acts like every project is a greenfield tutorial.\u003C\u002Fli>\u003C\u002Ful>\u003Cp>How to apply it: give the agent your ugliest real task, not a clean toy problem. Use a repo with existing conventions and an annoying edge case. Then measure whether it preserves the project’s shape or bulldozes it.\u003C\u002Fp>\u003Ch2>What I’d want before I trust Grok Build in a real workflow\u003C\u002Fh2>\u003Cp>Before I let any coding agent near a serious branch, I want a few boring answers. What context does it actually use? How are changes reviewed? Can I inspect every step? Can I stop it halfway? Does it produce clean diffs, or does it scatter edits like confetti? These are not glamorous questions, but they’re the whole game.\u003C\u002Fp>\u003Cp>If Grok Build is going to matter, xAI needs to make the review loop obvious. Developers do not want magic here. We want traceability. I want to know why it changed a line. I want to know whether it checked tests or just guessed. I want to know whether it can recover when the first attempt fails, because it will fail. Every agent does.\u003C\u002Fp>\u003Cp>This is where a lot of AI coding products get lazy. They show the happy path and leave the rest to the user. That works until the first bad patch lands in a shared branch. Then the team stops trusting the tool, and once trust is gone, the product becomes shelfware.\u003C\u002Fp>\u003Cp>How to apply it: before adopting any agent, write down your non-negotiables. For me, that list is short: readable diffs, controllable scope, visible reasoning, and a clean way to reject bad changes. If a tool can’t meet those, I don’t care how clever it is.\u003C\u002Fp>\u003Ch2>How I’d test Grok Build without wasting a week\u003C\u002Fh2>\u003Cp>If I were evaluating Grok Build tomorrow, I wouldn’t start with a greenfield app. I’d start with a small but annoying task in a real codebase. Something like updating an API contract, fixing a flaky test, or moving a helper without breaking import paths. That tells me far more than “generate a calculator app.”\u003C\u002Fp>\u003Cp>I’d also compare it against tools people already use, like \u003Ca href=\"https:\u002F\u002Fopenai.com\u002Findex\u002Fintroducing-codex\u002F\">OpenAI Codex\u003C\u002Fa> and \u003Ca href=\"\u002Ftag\u002Fcopilot\">Copilot\u003C\u002Fa>. Not because the logos matter, but because the workflow does. Can Grok Build stay useful across multiple turns? Can it recover from a wrong turn? Can it explain tradeoffs without turning into a lecture?\u003C\u002Fp>\u003Cp>And I’d keep a close eye on whether it behaves like a collaborator or a vending machine. A vending machine gives you code on demand. A collaborator helps you avoid making a mess. That’s the standard I’d use, and it’s the only one that’s worth paying attention to.\u003C\u002Fp>\u003Cp>How to apply it: build a one-hour eval. Pick three tasks, each with a different kind of pain: one refactor, one bug fix, one test change. Score the agent on accuracy, diff quality, and how much cleanup you had to do afterward. If the cleanup is bigger than the help, you have your answer.\u003C\u002Fp>\u003Ch2>The template you can copy\u003C\u002Fh2>\u003Cpre>\u003Ccode># AI coding agent eval template\n\n## Goal\nDescribe the exact task you want the agent to complete.\n\n## Repo context\n- Language\u002Fframework:\n- Package manager:\n- Test command:\n- Files likely involved:\n\n## Task\nWrite the change request in one paragraph.\n\n## Constraints\n- Do not change public APIs unless required.\n- Keep diffs small.\n- Preserve existing style and naming.\n- Update tests if behavior changes.\n- Explain any tradeoffs before editing.\n\n## Questions the agent should answer first\n1. What files do you think are relevant?\n2. What risks do you see?\n3. What is the smallest safe change?\n\n## What good output looks like\n- Clear plan before edits\n- Small, reviewable diff\n- Tests updated or added\n- Notes on edge cases and follow-up work\n\n## Red flags\n- Rewrites unrelated files\n- Ignores existing conventions\n- Makes broad changes without asking\n- Produces code that looks right but is hard to review\n\n## Scorecard\n- Repo understanding: 1-5\n- Diff quality: 1-5\n- Scope control: 1-5\n- Test awareness: 1-5\n- Cleanup required: 1-5\n\n## Decision\n- Adopt\n- Pilot\n- Reject\n\n## Notes\nCapture the agent’s best and worst behaviors here.\n\u003C\u002Fcode>\u003C\u002Fpre>\u003Cp>That’s the version I’d actually use. Not a hype checklist, just a way to see whether the agent can survive contact with a real codebase.\u003C\u002Fp>\u003Cp>Source attribution: I’m breaking down PCMag’s report on xAI’s Grok Build launch at \u003Ca href=\"https:\u002F\u002Fwww.pcmag.com\u002Fnews\u002Felon-musks-xai-launches-grok-build-its-first-ai-coding-agent\">https:\u002F\u002Fwww.pcmag.com\u002Fnews\u002Felon-musks-xai-launches-grok-build-its-first-ai-coding-agent\u003C\u002Fa>. The framing and template here are my own; the underlying product announcement is xAI’s.\u003C\u002Fp>","Grok Build is xAI’s first coding agent, and I break down how to use the idea without the usual agent fluff.","www.pcmag.com","https:\u002F\u002Fwww.pcmag.com\u002Fnews\u002Felon-musks-xai-launches-grok-build-its-first-ai-coding-agent",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779174862471-tn1e.png",[13,14,15,16,17],"xAI","Grok Build","AI coding agents","developer workflow","agent eval","en",0,false,"2026-05-19T07:13:53.995944+00:00","2026-05-19T07:13:53.986+00:00","done","6f0990a4-a816-4227-8a7e-29c851ececa7","grok-build-turns-xai-into-coding-agent-en","ai-agent","3ee7a75a-d5c8-4b02-98aa-c37375a97efc","published","2026-05-19T09:00:32.678+00:00",[31,32,33],"Treat Grok Build as a workflow claim, not just a model launch.","Test whether the agent can handle real repo constraints and push back on bad plans.","Use a small eval template to judge diff quality, scope control, and cleanup cost.","a9bee732-b07c-4e5b-a0e6-3048577e32a7","[0.010902193,0.011663357,0.013038594,-0.103843816,-0.012097087,-0.003672907,-0.016711583,0.0007837143,-0.0072741555,-0.0038993058,-0.011379213,0.02372235,0.025541635,0.018726734,0.14346442,0.022522662,-0.012493927,0.0063443827,-0.009410208,-0.008056946,-0.001261438,0.023092821,-0.004767678,-0.017790006,0.0023121852,-0.013537838,0.007372775,0.011285074,0.042246193,0.005383753,-0.017402926,0.02402581,-0.022398088,-0.009593365,0.015952343,0.02723363,0.022953246,-0.006848423,-0.005788856,-0.030901173,-0.016186224,-0.0079738265,-0.006496255,0.0013164593,-0.0021898514,0.008312679,-0.016470471,-0.0058200853,-0.010326133,0.030261451,-0.007062383,0.025918357,-0.00031726478,-0.15456618,0.029778093,0.008888466,-0.003696128,0.0054947967,0.023672916,0.0038264373,-0.024940351,0.0075723976,-0.013422572,-0.025339458,0.0045186575,-0.011667512,0.024202127,-0.004694887,-0.0057745785,-0.006879006,-0.0011498712,-0.00095756195,0.010492076,-0.024732497,-0.0110090105,-0.023890346,-0.007720744,0.018843865,0.019260764,0.0243049,0.013289422,-0.004762794,-0.005649288,-0.013945491,-0.00085924065,-0.015045049,-0.0005571816,-0.0046592876,0.006099314,-0.017762842,-0.012504732,0.0074566435,0.011427937,0.00089728425,0.005075327,-0.008410117,0.012206271,-0.013632514,-0.0085466765,0.007027752,-0.008592645,-0.019415632,-0.021003064,-0.019357625,0.012036209,0.012581191,-0.007812039,0.009761467,-0.0053836657,0.010701093,0.024656123,-0.010647492,-0.0033379898,-0.009791037,0.019599954,-0.11782963,0.02944454,-0.0027629018,0.0069632544,0.012753139,-0.02560438,0.019211689,-0.02135019,0.022896077,-0.0032496112,0.016250907,-0.019411972,0.023870638,-0.02651597,0.011181539,-0.020765547,-0.012844473,0.004297056,-0.006735762,0.023108667,0.006404687,-0.011112338,-0.007011076,-0.033178613,-0.011930794,-0.028707752,0.027836576,0.009727452,-0.013277672,-0.020976996,-0.014518801,-0.016472688,0.022341253,-0.0075627435,-0.021649456,0.011353211,-0.02333022,-0.008302992,-0.0057626916,0.0053794254,-0.05851556,0.008050577,-0.010618012,0.0018084082,0.0006290732,-0.011985017,-0.023110459,0.0042450223,0.026620919,-0.011103616,-0.01682995,-0.0073200897,-0.022999108,-0.00054916507,0.01685479,0.010421926,-0.010963762,0.004182585,0.012978909,0.018592672,-0.00966921,0.004245885,-0.0022430932,0.00023658125,-0.014227291,-7.4861964e-05,0.008534319,-0.015119269,0.011531585,0.0073510953,0.010314361,-0.02281371,0.006649448,0.018029632,-0.016666368,-0.018850515,-0.0054106396,0.03400671,0.010383462,0.022497129,-0.020459596,-0.0076378104,0.0150643345,0.011880924,0.03022803,0.026089605,-0.0116139315,-0.001336473,-0.032352347,-0.003255283,0.015620477,0.0028874627,-0.021956474,-0.012147398,-0.020757401,-0.016261835,0.013507512,0.025749631,-0.0064668246,0.011217786,0.002564678,-0.0043441765,-0.000870898,0.011408103,-0.0013891832,0.010238962,-0.012786472,0.030727282,0.016801633,-0.014104318,-0.023813108,-0.015061643,-0.008775407,-0.01878477,0.02909887,0.014654727,0.0143974535,-0.017249314,0.00551259,-0.0111329425,0.03478375,0.009449108,-0.015174392,-0.010015835,0.036134526,-0.023963718,-0.0035545663,-0.006483674,0.026973031,0.050972637,-0.01808774,0.02740934,-0.00660937,-0.025505692,-0.010907895,-0.025172072,0.027265767,-0.030205222,-0.038133387,-0.00812755,-0.0030074578,-0.0034059386,0.024819395,-0.0057527665,-0.0034232682,-0.025153168,-0.001292911,-0.041971233,-0.0011570373,0.005233552,0.0043294835,0.010981675,0.01404038,-0.021555685,0.006884875,-0.013566524,0.008027789,-0.013631825,-0.008877977,-0.0014737911,0.02596008,-0.045410413,0.036160782,0.008429708,-0.02111578,-0.016314736,0.006302389,-0.009694896,0.002589779,-0.0038123222,0.0052112397,-0.024209976,-0.016181022,0.0027269474,-0.012202534,-0.021063827,0.026979545,-0.009995596,-0.0057057976,-0.008778516,-0.026068257,-0.0014829754,-0.0019976585,0.013498998,0.01663715,0.031762972,-0.029576838,0.008470132,0.022696486,-0.040559094,-0.004548946,0.018149558,0.006328538,-0.0071809567,0.004508471,-0.004057533,-0.000469589,-0.013843928,-0.03494982,0.00699149,0.007184078,0.0032041103,-0.013036546,0.022517815,-0.024831912,-0.0060533606,-0.0033063574,0.004228348,-0.016744068,-0.017748784,-0.007885157,0.017662043,-0.01490592,0.009942407,-0.011800162,0.009149452,0.014457993,0.03673313,-0.052722476,0.0074222228,0.013052151,0.007349414,-0.025240308,-0.023414658,0.0032869433,0.015760856,-0.022530317,-0.011413175,-0.027967222,-0.032215606,0.032864008,0.003805641,-0.0018908618,0.011254878,-0.033741605,0.004560706,-0.012942931,-0.014366472,0.0022067763,-0.030840738,0.020349698,0.0062056286,-0.014122067,0.02203674,0.013274565,0.013733061,-0.018173387,-0.017406214,-0.0055119703,-0.0034777976,0.0009950787,0.03215094,0.03609802,-0.007482615,0.009569562,-0.004032374,-0.0009935669,0.012073155,-0.006511621,0.0013423716,-0.0083578285,-0.041027434,0.0010772836,-0.0084654875,0.013326721,-0.006933812,0.016989458,-0.0054885754,-0.026495084,-0.024730964,0.019133933,0.010391474,-0.00456736,0.0003140845,-0.001963234,-0.0031943114,-0.0089589115,-0.019924784,0.028617062,0.033943877,0.014216007,-0.008550367,0.03253798,-0.03458274,-0.0060823634,-0.009452838,0.016776899,-0.014178738,0.03194451,-0.03586919,0.0035036963,-0.013035514,-0.0002462121,0.017954953,0.017590286,-0.001618239,0.0025116596,0.031402566,0.010649133,-0.0002735117,-0.0069439653,-0.01057576,0.009597825,-0.004526617,-0.0017601908,0.0045726374,0.031128233,0.017152924,0.0022626389,-0.011041331,-0.01638339,-0.009710885,-0.009872918,0.0007578148,-0.0015594794,-0.009035999,-0.0042536356,0.023900049,-0.02052302,-0.035876397,-0.012640659,-0.023971522,-0.0032960186,-0.03240674,-0.007746651,-0.00051484705,-0.014834405,0.003989602,0.03969958,0.009298017,0.004584845,0.0238773,-0.009199555,-0.050833497,0.01860165,0.0073517924,0.017081903,0.01969376,-0.011550988,0.008345239,0.013563857,0.0024356327,0.02137509,0.0050882744,-0.0077538076,0.032457594,0.0068883887,0.0066222567,0.009935348,-0.01709415,-0.0075271097,0.0015643104,-0.023605444,-0.022391917,0.012081891,-0.011448282,0.022579027,-0.0048331697,-0.031094605,-0.026851417,0.018251512,-0.031907395,-0.012886164,0.0050598825,0.013599147,0.0073247175,0.027734214,0.02768197,0.00082952104,-0.029961629,0.016599378,-0.0024802582,-0.024116183,-0.00052393985,-0.0029948712,0.052976463,0.024452459,0.011222828,-0.013922071,-0.04269095,0.00084353535,-0.011089256,-0.016046599,-0.014293246,-0.0033555897,-0.01828833,-0.008353478,0.00558095,-0.00086915365,0.0006388108,-0.019324604,-0.023701804,-0.019115895,0.0053556818,0.0065445895,-0.01664788,-0.0021057618,0.0046779905,0.005743059,-0.0012685846,-0.02331295,-0.0055841613,-0.009624946,0.020226896,-0.0023818454,-0.0059078303,0.011556911,-0.017987793,0.003521623,-0.02405858,-0.010514995,0.027987028,0.014766816,0.013323733,0.008311623,-0.005326404,0.009011641,-0.009513792,0.00033997901,0.003240361,0.017565198,0.014921627,0.0034918217,-0.002217539,0.016772367,-0.04219346,-0.012732234,-0.01830565,-0.0076133558,0.014113908,-0.11503858,0.020827733,-0.007825806,-9.071745e-05,-0.0022741167,0.010341558,-0.013571742,-0.023393707,-0.0038334387,-0.008691143,-0.0037963162,-0.014951779,0.027994879,0.0075206696,-0.0029427228,0.009924802,0.009771849,-0.010744619,0.023968423,-0.032498877,0.031085309,-0.0033957,0.0048677726,-0.014233999,0.0034097724,-0.002690004,0.025633477,0.004437016,0.0033658948,-0.01460426,-0.022891352,-0.006960955,0.016380103,0.019359658,0.005118653,0.012703349,0.010378187,-7.391271e-05,0.008687647,0.006823138,0.02411842,0.013141608,0.00022330174,-0.0017844933,-0.013949177,0.025773974,0.03687122,-0.016690906,0.009171745,0.02123983,-0.060517635,-0.006382265,-0.019530658,-0.024774851,0.0008992008,0.02216055,-0.02655378,-0.0045841946,0.0066661555,-0.0016713121,0.013273732,-0.009739656,-0.00023832482,0.008185117,-0.0005337814,0.024075069,0.0071128835,0.042737853,0.00061526964,0.0032297324,0.011674615,-0.034956165,0.014864656,0.014632079,0.013288251,-0.002015801,-0.009656368,0.011420166,-0.0042321226,0.022848433,-0.04059184,-0.032174844,-0.062213697,-0.022994189,0.0062310565,-0.014381599,-0.0023168328,0.0037893986,-0.006287113,-0.018325502,-0.022912502,0.022070793,0.011649057,-0.005682006,-0.020498557,-0.005871436,-0.015586306,0.019827327,-0.018114699,-0.010479858,0.0040207524,-0.017217403,-0.017850008,0.0076630497,0.010244537,-0.025389696,0.0039816,0.04376053,-0.03155965,-0.0066872747,-0.0082174055,-0.019916901,-0.007294227,-0.11461332,0.00658118,0.003768052,0.005333603,-0.0013540188,0.005674307,0.014944866,0.0121664135,0.006393792,-0.018883256,0.007240094,-0.015188511,-0.028005196,-0.017776063,0.012952315,0.13475852,-0.036151085,-0.044122703,-0.011032371,0.018404104,0.015251157,-0.03516778,-0.023824532,0.017399253,-0.011230774,-0.013536817,0.013491158,-0.03038561,0.02472412,0.0027859726,0.006144756,0.024622317,-0.0067453757,-0.013472737,0.019341651,-0.0043881866,0.003265846,-0.011826256,0.009619084,0.03475913,0.0039661545,0.012839628,0.018886518,-0.007109409,0.009323225,0.0021553577,-0.027277652,-0.005887939,-0.024505608,-0.012216796,-0.0030004554,-0.073995836,0.016979367,-0.022080008,-0.006684912,0.012192556,0.0070754783,0.022073217,-0.005611249,0.02353498,0.021250222,0.005045219,-0.0017695107,0.037625577,0.0056778835,0.0073460517,0.0068752393,0.016975058,0.00844141,0.007898014,-0.018329035,-0.0024561463,-0.0074873776,0.038254987,0.0013978498,-0.013010235,0.0016931397,0.023469038,0.014849027,0.0061728302,0.0037482511,-0.027940892,0.02365918,-0.0025309932,0.013990892,0.017995875,0.0040648994,-0.02199683,0.010662117,-0.018816113,-0.01914514,0.010488978,0.0015041293,0.025074253,-0.0032615974,0.053875178,-0.010851695,0.016767832,0.007187795,0.018703086,0.006967035,-0.018782007,0.015919216,-0.026869161,-0.008583783,0.015130309,0.029320423,0.022606034,0.017991219,0.025267946]",{"tags":37,"relatedLang":48,"relatedPosts":52},[38,40,42,44,46],{"name":16,"slug":39},"developer-workflow",{"name":13,"slug":41},"xai",{"name":17,"slug":43},"agent-eval",{"name":14,"slug":45},"grok-build",{"name":15,"slug":47},"ai-coding-agents",{"id":27,"slug":49,"title":50,"language":51},"grok-build-turns-xai-into-coding-agent-zh","Grok Build 把 xAI 變成寫碼代理","zh",[53,59,65,71,77,83],{"id":54,"slug":55,"title":56,"cover_image":57,"image_url":57,"created_at":58,"category":26},"3a352268-d5bf-423e-98dc-b3fbfe4add43","agentic-ai-turns-autonomy-into-security-problem-en","Agentic AI turns autonomy into a security problem","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779139471362-dv7k.png","2026-05-18T21:24:05.681031+00:00",{"id":60,"slug":61,"title":62,"cover_image":63,"image_url":63,"created_at":64,"category":26},"a2a8f68d-754a-41d9-a85a-8564d8106be3","why-googles-gemini-spark-should-worry-anyone-using-ai-agents-en","Why Google’s Gemini Spark should worry anyone using AI agents","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779138846928-23xx.png","2026-05-18T21:13:30.005705+00:00",{"id":66,"slug":67,"title":68,"cover_image":69,"image_url":69,"created_at":70,"category":26},"e839503a-27d2-4164-813d-e1f6891c477d","microsoft-copilot-2026-update-real-workflows-en","Microsoft Copilot’s 2026 update targets real work","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779132246480-du2f.png","2026-05-18T19:23:37.494277+00:00",{"id":72,"slug":73,"title":74,"cover_image":75,"image_url":75,"created_at":76,"category":26},"d558c492-a87a-4bb5-883c-3f62b726fbce","how-to-engineer-prompts-for-ai-agents-en","How to Engineer Prompts for AI Agents","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779131631690-tifb.png","2026-05-18T19:13:23.476836+00:00",{"id":78,"slug":79,"title":80,"cover_image":81,"image_url":81,"created_at":82,"category":26},"b94a8bd9-9152-47f8-816d-4b543a623941","why-browser-agents-need-real-execution-layer-en","Why browser agents need a real execution layer, not another wrapper","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779066232000-ktdk.png","2026-05-18T01:03:24.666344+00:00",{"id":84,"slug":85,"title":86,"cover_image":87,"image_url":87,"created_at":88,"category":26},"e8a5de94-8a32-4343-a44d-d6c8d10c8bbc","why-openai-is-right-to-put-codex-on-phones-en","Why OpenAI Is Right to Put Codex on Phones","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779019443877-qwlw.png","2026-05-17T12:03:35.468708+00:00",[90,95,100,105,110,115,120,125,130,135],{"id":91,"slug":92,"title":93,"created_at":94},"03db8de8-8dc2-4ac1-9cf7-898782efbb1f","anthropic-claude-ai-agent-task-automation-en","Anthropic's Claude AI Agent: A New Era of Task Automation","2026-03-25T16:25:06.513026+00:00",{"id":96,"slug":97,"title":98,"created_at":99},"045d1abc-190d-4594-8c95-91e2a26f0c5a","googles-2026-ai-agent-report-decoded-en","Google’s 2026 AI Agent Report, Decoded","2026-03-26T11:15:23.046616+00:00",{"id":101,"slug":102,"title":103,"created_at":104},"e64aba21-254b-4f93-aa21-837484bb52ec","kimi-k25-review-stronger-still-not-legend-en","Kimi K2.5 review: stronger, still not a legend","2026-03-27T07:15:55.385951+00:00",{"id":106,"slug":107,"title":108,"created_at":109},"30dfb781-a1b2-4add-aebe-b3df40247c37","claude-code-controls-mac-desktop-en","Claude Code now controls your Mac desktop","2026-03-28T03:01:59.384091+00:00",{"id":111,"slug":112,"title":113,"created_at":114},"254405b6-7833-4800-8e13-f5196deefbe6","cloudflare-100x-faster-ai-agent-sandbox-en","Cloudflare’s 100x Faster AI Agent Sandbox","2026-03-28T03:09:44.356437+00:00",{"id":116,"slug":117,"title":118,"created_at":119},"04f29b7f-9b91-4306-89a7-97d725e6e1ba","openai-backs-isara-agent-swarm-bet-en","OpenAI backs Isara’s agent-swarm bet","2026-03-28T03:15:27.849766+00:00",{"id":121,"slug":122,"title":123,"created_at":124},"3b0bf479-e4ae-4703-9666-721a7e0cdb91","openai-plan-automated-ai-researcher-en","OpenAI’s plan for an automated AI researcher","2026-03-28T03:17:42.312819+00:00",{"id":126,"slug":127,"title":128,"created_at":129},"fe91bce0-b85d-4efa-a207-24ae9939c29f","harness-engineering-ai-agent-reliability-2026","Harness Engineering: From Bridle to Operating System, The Missing Link in AI Agent Reliability","2026-03-31T06:36:55.648751+00:00",{"id":131,"slug":132,"title":133,"created_at":134},"67dc66da-ca46-4aa5-970b-e997a39fe109","openai-codex-plugin-claude-code-en","OpenAI puts Codex inside Claude Code","2026-04-01T09:21:55.381386+00:00",{"id":136,"slug":137,"title":138,"created_at":139},"7a09007d-820f-43b3-8607-8ad1bfcb94c8","mcp-explained-from-prompts-to-production-en","MCP Explained: From Prompts to Production","2026-04-01T09:24:40.089177+00:00"]