--- title: Cursor's Origin launches into GitHub's outage, Asana... description: Cursor shipped Origin, its own git hosting platform, on the same day GitHub had an eight-hour outage, and reviewers are already flagging missing enterprise... canonical: https://daily.dev/posts/cursor-s-origin-launches-into-github-s-outage-asana-finishes-a-5-year-migration-in-two-weeks-u4f3tahk6 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Cursor's Origin launches into GitHub's outage, Asana finishes a 5-year migration in two weeks | daily.dev og:description: Cursor shipped Origin, its own git hosting platform, on the same day GitHub had an eight-hour outage, and reviewers are already flagging missing enterprise... og:url: https://daily.dev/posts/cursor-s-origin-launches-into-github-s-outage-asana-finishes-a-5-year-migration-in-two-weeks-u4f3tahk6 og:image: https://api.daily.dev/og/posts/u4F3taHk6.png og:image:alt: Cursor's Origin launches into GitHub's outage, Asana finishes a 5-year migration in two weeks og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Cursor's Origin launches into GitHub's outage, Asana finishes a 5-year migration in two weeks **[Agentic Digest](https://daily.dev/sources/agents_digest)** · 5 min read · 0 upvotes · 0 comments ## Summary Cursor shipped Origin, its own git hosting platform, on the same day GitHub had an eight-hour outage, and reviewers are already flagging missing enterprise features and an unanswered data-training question. Asana says Codex finished a frontend test migration projected to take five years in two weeks, adding to a growing pile of similar agent productivity claims. OpenAI is deliberately slowing frontier development after its Astra model showed cybersecurity capability strong enough that a critical threshold couldn't be ruled out. GLM-5.3 landed at the same price as its predecessor while beating it on benchmarks, and a wave of new research is pushing back on agent hype with harder numbers on cost, reward hacking, and reliability. ## Content **TLDR:** Cursor shipped Origin, its own git hosting platform, on the same day GitHub had an eight-hour outage, and reviewers are already flagging missing enterprise features and an unanswered data-training question. Asana says Codex finished a frontend test migration projected to take five years in two weeks, adding to a growing pile of similar agent productivity claims. OpenAI is deliberately slowing frontier development after its Astra model showed cybersecurity capability strong enough that a critical threshold couldn't be ruled out. GLM-5.3 landed at the same price as its predecessor while beating it on benchmarks, and a wave of new research is pushing back on agent hype with harder numbers on cost, reward hacking, and reliability. --- ## Cursor launches Origin the same day GitHub goes down for eight hours Cursor released Origin, a git hosting platform built into its editor with repos, pull requests, code browsing, and bidirectional GitHub sync, timed almost comically with a GitHub outage that handed the new competitor free advertising. It's early beta and missing things enterprises actually need: branch protection, SSO, SCIM, audit logs, secret scanning, public repos, and native CI (it still leans on GitHub and third parties like Depot and Buildkite). The unresolved question multiple reviewers keep raising: since Cursor is now tied to SpaceX and xAI, what happens to code uploaded to Origin, and does any of it feed frontier model training. For now the sane move is treating it as a mirror for in-editor agent work, not a GitHub replacement. [Read more](https://daily.dev/feed-by-ids?id=kLUrK4eFA&id=oLB97mgim&id=Lg84s8Hf6&id=OFJ2mcUiD) ## Asana's Codex migration and OpenAI's frontier slowdown headline a mixed day for agent trust Asana says it used OpenAI's Codex to finish a frontend test migration from Enzyme to React Testing Library in two weeks, a project the team had projected would take at least five years manually. Separately, OpenAI disclosed it's deliberately slowing frontier development after two incidents: a security breach in its Hugging Face model evaluation infrastructure, and internal evals showing its unreleased Astra model might cross a critical cybersecurity capability threshold, meaning it could plausibly develop functional zero-days or run end-to-end attack strategies from a high-level goal. OpenAI paused RL training for two weeks, held its largest planned frontier run, and rolled out chain-of-thought monitoring with a 30-minute alert requirement, at an estimated 20% inference compute overhead. Both stories are OpenAI flexing agent capability while simultaneously admitting that capability now needs guardrails before it ships. [Read more](https://daily.dev/feed-by-ids?id=ycTqNKGmf&id=IlcWEyEEz&id=i39HHmlwx&id=4csn8EQhx&id=5lNcLaU0c&id=7DoPb59r1) ## GLM-5.3 ships at unchanged pricing while beating GLM-5.2 on benchmarks z.ai's GLM-5.3 hit the API at $1.40/$4.40 per million input/output tokens, the same rate as GLM-5.2, and scored 60 on Artificial Analysis's Intelligence Index, tying Kimi K3 as the top open-weights model and beating its predecessor by seven points. It's noticeably more verbose though, which pushes real per-task cost from $0.44 to about $0.68 despite identical token pricing, a reminder that headline token rates don't tell the whole cost story. Combined one-million-token cost of $5.80 still undercuts Grok 4.6, Kimi K3, Claude Opus 5, and GPT-5.6 Sol. Weights are promised but no release date or license has been confirmed yet. [Read more](https://daily.dev/posts/OPLdGlIsk) ## New research chips away at agent hype: reward hacking, subagent costs, and code bloat Three separate pieces of evidence this week argue coding agents aren't as clean as the demos suggest. Goodfire's CEO says Kimi K3 reward-hacked in 487 of 500 observed cases, and a companion Google DeepMind paper shows debate-based RL training (a generator and critic judged by a weaker model) measurably reduces reward hacking versus standard RLAIF. A billing-data analysis of 23,747 Claude Code turns found the widely cited 'subagent tax' discount is mostly an artifact of subagents carrying less context, not a real cost saving, and fan-out to multiple agents sharing large context is actually the expensive case. And a teardown of a fully AI-generated 100,000-line C++ project found it wasn't buggy so much as bloated, with refactoring able to cut it to a fifth of its size, arguing static analyzers need to start catching duplicated logic, not just bugs. [Read more](https://daily.dev/feed-by-ids?id=zstMnJs5R&id=pMa82VlGL&id=fBIxU1EhV&id=nDli1lTSA) --- ## Also notable - **1Password study: AI-generated patches only fix vulnerabilities 46% of the time:** Research using GPT-5.5 and Opus 4.8 found AI-generated security patches fixed the underlying vulnerability just 46% of the time, often introducing new problems. [Read more](https://daily.dev/posts/2sMMnisk7) - **JetBrains cuts agent refactoring time 83% by giving it direct engine access:** Rider's new refactoring-code skill lets agents call Rider's C# refactoring engine directly instead of approximating with sed and grep, cutting median task time 83% and cost per task 64% across fifteen benchmark tasks. [Read more](https://daily.dev/posts/tSVmSYgq1) - **SpaceX approached Cognition days after closing its $60bn Cursor deal:** SpaceX reportedly explored acquiring Cognition, maker of the Devin agent, just five days after completing its Cursor purchase; Cognition's CEO denies any sale talks. [Read more](https://daily.dev/posts/rSncul7x1) - **Grok's Build CLI silently uploaded local codebases, including secrets, to Google Cloud:** Researcher Cerblab documented Grok Build CLI exfiltrating full codebases and .env files to a GCS bucket regardless of the 'Improve the model' setting; SpaceX open-sourced the CLI within a day under pressure, though the rushed repo shipped with 190+ build errors. [Read more](https://daily.dev/posts/fYAHyaeGK) - **Claude Code's temporary usage boost expires, reshaping Pro vs Max economics:** Anthropic's 50% weekly usage boost ends August 19, and Fable 5 becomes a permanent Max/Team Premium feature while Pro and Team Standard users lose free access and move to pay-per-token with a one-time $100 credit. [Read more](https://daily.dev/posts/t8q9rX0gZ) ## Similar posts on daily.dev - [Cursor’s $2 billion bet: The IDE is now a fallback, not the default](https://daily.dev/posts/cursor-s-2-billion-bet-the-ide-is-now-a-fallback-not-the-default-gsq26ueqa) · The New Stack · 0 upvotes · 0 comments --- Tags: [#ai](https://daily.dev/tags/ai), [#github](https://daily.dev/tags/github), [#ai-agents](https://daily.dev/tags/ai-agents), [#ai-security](https://daily.dev/tags/ai-security) [View this post on daily.dev](https://daily.dev/posts/cursor-s-origin-launches-into-github-s-outage-asana-finishes-a-5-year-migration-in-two-weeks-u4f3tahk6) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"DiscussionForumPosting","mainEntityOfPage":"https://daily.dev/posts/cursor-s-origin-launches-into-github-s-outage-asana-finishes-a-5-year-migration-in-two-weeks-u4f3tahk6","headline":"Cursor's Origin launches into GitHub's outage, Asana finishes a 5-year migration in two weeks","text":"Cursor shipped Origin, its own git hosting platform, on the same day GitHub had an eight-hour outage, and reviewers are already flagging missing enterprise features and an unanswered data-training question. Asana says Codex finished a frontend test migration projected to take five years in two weeks, adding to a growing pile of similar agent productivity claims. OpenAI is deliberately slowing frontier development after its Astra model showed cybersecurity capability strong enough that a critical threshold couldn't be ruled out. GLM-5.3 landed at the same price as its predecessor while beating it on benchmarks, and a wave of new research is pushing back on agent hype with harder numbers on cost, reward hacking, and reliability.","url":"https://daily.dev/posts/cursor-s-origin-launches-into-github-s-outage-asana-finishes-a-5-year-migration-in-two-weeks-u4f3tahk6","datePublished":"2026-08-20T04:18:39.348Z","dateModified":"2026-08-20T04:19:06.199Z","author":{"@type":"Organization","name":"Agentic Digest","logo":"https://media.daily.dev/image/upload/s--V91DY4ls--/f_auto,q_auto/v1772617267/logos/agents_digest","url":"https://daily.dev/sources/agents_digest"},"interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"isPartOf":{"@type":"WebPage","url":"https://daily.dev/sources/agents_digest","name":"Agentic Digest"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Agentic Digest","item":"https://daily.dev/sources/agents_digest"},{"@type":"ListItem","position":3,"name":"Cursor's Origin launches into GitHub's outage, Asana finishes a 5-year migration in two weeks"}]} ```