{"rewrite":{"id":"r_a06ab462bb1055bba115a3b6","clusterId":"c_99dfb417342c7eaae5fe0a1c","slug":"brave-develops-agentstop-to-halt-wasted-local-ai-agent-runs","model":"deepseek-v4-flash","headline":"Brave Develops AgentStop to Halt Wasted Local AI Agent Runs","summary":"Brave, the company behind the privacy-focused browser, announced on May 28, 2026, a new system called AgentStop that detects when a local AI agent is failing a task and terminates the process early. Running AI locally consumes a device's own compute and battery, and agents performing complex tasks can loop for over ten minutes, calling the language model dozens of times, pushing GPU power past 40 watts, and keeping temperatures above 90 degrees Celsius for extended periods, often still failing the task. AgentStop monitors the agent's output for signs of failure, such as low confidence in output tokens, an abnormally high number of processing tokens per step indicating a loop, or repeated identical results. In a benchmark test using the Qwen3-Coder-30B-A3B model on 500 tasks from SWE-Bench Verified, enabling AgentStop reduced power consumption by about 19% while only lowering task completion rate by roughly 3%. Brave has released AgentStop as an open-source project under the MIT License on GitHub.","whyItMatters":"AgentStop tackles a growing practical barrier to local AI adoption by addressing the energy waste and hardware strain from runaway agent loops, making on-device AI more viable for everyday consumer laptops.","webCardHtml":"\u003cp\u003eLocal AI agents, unlike simple chat interfaces, can run extended inference chains that waste battery and GPU resources when they fail. Brave's AgentStop system watches the agent's output stream for three failure signals: low confidence in generated tokens, an excessive number of processing tokens per step that suggests the agent is stuck in a loop, and repeated output of the same result. When any of these patterns appear, AgentStop terminates the agent early.\u003c/p\u003e\u003cp\u003eBrave tested AgentStop with the Qwen3-Coder-30B-A3B model on a MacBook Pro with an M1 Max chip, running 500 tasks from the SWE-Bench Verified benchmark. With AgentStop active, power consumption dropped roughly 19 percent compared to running without it, while the task completion rate fell only about 3 percent. The company positions AgentStop as a first step toward making local AI agents energy-efficient as well as private and convenient.\u003c/p\u003e\u003cp\u003eThe source code is available on GitHub under the MIT License. Brave's announcement was published on May 28, 2026.\u003c/p\u003e","blueskyPost":"Brave's AgentStop accepts a 3% task completion reduction to cut power use by 19%. The tradeoff is explicit: local AI efficiency now has a calibrated knob, not a kill switch.","twitterPost":"AgentStop trades a 3% task completion drop for 19% power savings. Brave made the calculus public.","threadsPost":"Brave's AgentStop monitors local AI agents for looping or low confidence and kills failing tasks early. The benchmark numbers are specific: 19% less power for a 3% completion hit. That ratio is the design choice, not the tech. Brave shipped the tradeoff as open source under MIT.","newsletterBlurb":"Brave announced AgentStop, a system that monitors local AI agents for signs of failure and terminates them early to save battery and compute. In benchmarks, it cut power consumption by about 19% while barely affecting task completion rates. The code is open-source under MIT.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20260529-agentstop-terminating-local-ai-brave/\",\"title\":\"「ローカルAIがズルズルと動き続けて無駄にバッテリーやGPUリソースを消費してしまう問題」を解決する技術「AgentStop」がBraveによって開発される\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":5422,"outputTokens":763,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1780204231,"createdAt":"2026-05-31T04:58:47.000Z","publishedAt":"2026-05-31T05:02:29.000Z","updatedAt":"2026-05-31T05:02:29.000Z"},"cluster":{"id":"c_99dfb417342c7eaae5fe0a1c","canonicalTitle":"「ローカルAIがズルズルと動き続けて無駄にバッテリーやGPUリソースを消費してしまう問題」を解決する技術「AgentStop」がBraveによって開発される","representativeArticleId":"a_185d87232f40e77d345859d4","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[],\"people\":[],\"type\":\"news\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-05-29T03:11:00.000Z","lastSeenAt":"2026-05-29T03:11:00.000Z","updatedAt":"2026-05-31T05:02:29.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20260529-agentstop-terminating-local-ai-brave/","title":"「ローカルAIがズルズルと動き続けて無駄にバッテリーやGPUリソースを消費してしまう問題」を解決する技術「AgentStop」がBraveによって開発される"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":[],"people":[],"type":"news","domain":"other","is_roundup":false},"keyFacts":["Brave announced AgentStop on May 28, 2026, a system that detects when a local AI agent is failing a task and terminates the process early.","Running AI locally can push GPU power past 40 watts and keep temperatures above 90 degrees Celsius for extended periods during failed agent runs.","AgentStop monitors for three failure signals: low confidence in output tokens, an abnormally high number of processing tokens per step, and repeated identical results.","In a benchmark test with the Qwen3-Coder-30B-A3B model on 500 tasks from SWE-Bench Verified, AgentStop reduced power consumption by about 19% while lowering task completion rate by roughly 3%.","Brave released AgentStop as an open-source project under the MIT License on GitHub."]}