[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-researchers-build-a-testbed-for-ai-agents-that-learn-over-time":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},9244,"researchers-build-a-testbed-for-ai-agents-that-learn-over-time","Researchers Build a Testbed for AI Agents That Learn Over Time","SCLATE is a new open scheduler that compresses month-long AI agent benchmarks into hours and shows bolted-on memory often loses to an agent's own memory.","A new research platform called SCLATE lets AI agents and the benchmarks testing them share one clock, cutting month-long evaluations down to hours.\n\nResearchers built SCLATE as a shared event scheduler so agent harnesses, memory systems, and benchmarks all plug into one timeline instead of each pairing writing its own scheduling code. A hybrid simulated clock runs in real time while an agent is actively working, then fast-forwards through idle stretches, compressing a month of simulated activity into a few hours of wall-clock time. The system also doubles as a rollout engine, running any agent's harness and memory setup unmodified while logging the tokens and log probabilities behind every model call through an in-container proxy. The team used it to port seven existing benchmarks and ran ten harness-and-memory combinations across ten different models.\n\nThe head-to-head results undercut a common assumption: bolting a fancier memory system onto an agent did not reliably beat whatever memory the harness already had built in, and different models got very different mileage out of the identical setup. That is a useful reality check for anyone buying a standalone memory layer on faith. The researchers also post-trained Qwen3.5-4B using unmodified harnesses and memory systems, and the model learned to use them better, reading 6.8 times fewer file lines while scoring 16.7 points higher on SWE-bench Verified, and gaining up to 11.8 points on a held-out benchmark called MetaClaw.\n\nIt is a plumbing paper, not a flashy one, but plumbing is exactly what has been missing before anyone can make real claims about which memory tricks actually help an agent remember.","[\"ai agents\",\"continual learning\",\"benchmarks\",\"qwen3.5\"]","2026-10-01T04:00:00.000Z","2026-10-02T04:21:13.287Z","2026-10-02T04:21:16.743Z","published",null,[],"ai",[26,27,28,29],"ai agents","continual learning","benchmarks","qwen3.5",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.32391",0,{"sections":36},[37,40,44,48,53,58,62,67,72,76,81,86,91,96],{"name":38,"slug":24,"count":39,"latest_published_at":18},"AI",5659,{"name":41,"slug":42,"count":43,"latest_published_at":18},"Security","security",818,{"name":45,"slug":46,"count":47,"latest_published_at":18},"Policy","policy",430,{"name":49,"slug":50,"count":51,"latest_published_at":52},"Deals","deals",298,"2026-09-30T21:00:26.000Z",{"name":54,"slug":55,"count":56,"latest_published_at":57},"Hardware","hardware",196,"2026-09-30T13:00:00.000Z",{"name":59,"slug":60,"count":61,"latest_published_at":18},"Science","science",163,{"name":63,"slug":64,"count":65,"latest_published_at":66},"Consumer Tech","consumer-tech",149,"2026-09-30T22:57:11.000Z",{"name":68,"slug":69,"count":70,"latest_published_at":71},"Dev Tools","dev-tools",93,"2026-10-01T02:30:48.000Z",{"name":73,"slug":74,"count":70,"latest_published_at":75},"Software","software","2026-09-30T21:41:11.000Z",{"name":77,"slug":78,"count":79,"latest_published_at":80},"Startups","startups",84,"2026-09-30T20:39:09.000Z",{"name":82,"slug":83,"count":84,"latest_published_at":85},"Gaming","gaming",51,"2026-09-30T16:24:30.000Z",{"name":87,"slug":88,"count":89,"latest_published_at":90},"General","general",50,"2026-09-30T21:37:54.000Z",{"name":92,"slug":93,"count":94,"latest_published_at":95},"Reviews","reviews",31,"2026-09-28T14:31:34.000Z",{"name":97,"slug":98,"count":99,"latest_published_at":100},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]