GitHub Enterprise Cloud customers with enterprise managed users can now access GitHub Copilot agent session data across all Copilot clients, including: Cloud agents operating on github.com and data resident deployments… The post Copilot agent session streaming is now in public preview appeared first on The GitHub Blog .
World Cup watch parties in Santa Clara and the San Francisco Bay Area brought fans together for the United States vs. Bosnia-Herzegovina match, where the U.S. earned a 2-0 win to reach the Round of 16.
Anthropic is reportedly in talks with Samsung Electronics about manufacturing a custom AI chip. The project is still early, but Anthropic has already hired chip engineers. After OpenAI's "Jalapeño," yet another major AI company is pushing into chip development to cut infrastructure costs. The article Anthropic reportedly explores custom chip manufacturing with Samsung while insisting Nvidia still matters appeared first on The Decoder .
Albanians have continued to protest against a planned luxury resort backed by Donald Trump's son-in-law, Jared Kushner. They say the resort threatens the coast's fragile ecosystems, including a flamingo habitat.
After its first year in office, Germany's coalition government is seeking to revive a sluggish economy and push through key changes with its broad package of reforms. The plans met with a mixed reaction.
Victims say hearing Keir Starmer’s acknowledgment of their suffering was emotional but want more mental support This summer’s World Cup fever vividly takes Ann Keen back to 1966 and the day England won the tournament – she was 17 years old and it was the day she told her father she was pregnant. “It was the worst thing that could ever have been said to him. I was told I’d put shame on the family and I must be sent away,” the former Labour MP said. “I was in an unmarried mother’s home where I had to scrub the steps from morning until night. It was all about punishment. Even in the delivery room I was told I couldn’t have anything for the pain, because I was a bad girl. And when NHS staff are telling you that, you start to believe it.” Continue reading...
I saw Geoffrey Litt speak at AIE yesterday, and one framing he used particularly resonated with me: Understand to participate Geoffrey was talking about the challenge of collaborating with coding agents as they construct increasingly large and sophisticated changes, and the need to avoid taking on cognitive debt as your understanding drifts from how the code actually works. His argument is that you need to understand the code to a depth that enables you to participate further with the model: You can learn what the agent is doing to make sure you can be an active participant in the creative process. [...] You need a rich set of concepts in your mind to think creatively and fluently about how to move something forward. If you're lacking that fluency, your ability to participate in the project is meaningfully limited. The AIE talks are all recorded - all 300+ of them! - and should be trickling out over the next three weeks. Geoffrey's is one that I recommend catching on YouTube. Geoffrey also published a thread version of his talk on Twitter. Tags: geoffrey-litt , coding-agents , cognitive-debt , generative-ai , ai , llms
NCA says offenders arrange to sexually assault and film victims via online networks with crimes often taking place in trusting relationships Criminal investigators in the UK say they have uncovered a “truly international network” of organised drug-facilitated sexual assault in which victims are sedated before being raped and sexually assaulted. The National Crime Agency [NCA] has said online networks, “many as yet unidentified by law enforcement”, were allowing offenders to arrange to rape and abuse victims or arrange for sexual assaults to be filmed. Continue reading...
When Subquadratic launched earlier this year, it could build a sparse-attention model that could handle a 12-million token context window The post What comes after attention? This startup says it already knows. appeared first on The New Stack .
As much of the Eastern U.S. experiences potentially record-breaking daytime temperatures, the nights are also staying unusually warm, leaving the human body no time to recover.
The NHS will divert billions of pounds from essential services to pay for new medicines, under the terms of the US-UK trade deal agreed in December, which could lead to more than 200,000 excess deaths, analysis has found. Ministers have defended the deal as a way of helping British drug exports avoid US tariffs and giving patients access to vital medication, but critics accuse the Labour party of caving into pressure from Donald Trump. Lucy Hough speaks to columnist Aditya Chakrabortty Continue reading...
Minor Changes #14474 aa5d580 Thanks @WillTaylorDev ! - Add cache options for WorkerEntrypoint exports You can now set cache options on WorkerEntrypoint exports and configure cross-version cache behavior globally: // wrangler.json { "cache" : { "enabled" : true , "cross_version_cache" : true }, "exports" : { "default" : { "type" : " worker " , "cache" : { "enabled" : false } }, "Admin" : { "type" : " worker " , "cache" : { "enabled" : true } } } } Wrangler sends the exports config to the deploy and version upload APIs alongside the global cache.enabled and cache.cross_version_cache settings. The platform resolves those global settings plus cache overrides on exports and validates which entrypoint names are cacheable. #14382 fd92d56 Thanks @petebacondarwin ! - Add support for declarative Durable Object exports wrangler deploy now accepts an exports map in wrangler.json as a declarative alternative to the legacy migrations array. Each entry in exports is keyed by Durable Object class name. type carries the export kind (currently always "durable-object" ); the state field carries the lifecycle and defaults to "created" (live) when omitted: { "exports" : { // Provision a new Durable Object class (`MyDO`) "MyDO" : { "type" : " durable-object " , "storage" : " sqlite " }, // Delete Durable Object class (`OldGone`) "OldGone" : { "type" : " durable-object " , "state" : " deleted " }, // Rename a Durable Object class (from `OldName` to `NewName`) "OldName" : { "type" : " durable-object " , "state" : " renamed " , "renamed_to" : " NewName " }, "NewName" : { "type" : " durable-object " , "storage" : " sqlite " }, // Transfer a Durable Object (`Outgoing`) to a new Worker (`target-worker`) "Outgoing" : { "type" : " durable-object " , "state" : " transferred " , "transferred_to" : " target-worker " }, // Prepare to receive the transfer of a Durable Object (`Incoming`) from another Worker (`source-worker`) "Incoming" : { "type" : " durable-object " , "state" : " expecting-transfer " , "storage" : " sqlite " , "transfer_from" : " source-worker " } } } When a Worker declares Durable Object class bindings but no lifecycle for them (neither a migrations array nor an exports map), wrangler warns and now suggests a declarative exports entry for each class (previously it suggested a legacy migrations block). The deployment response now surfaces the server's reconciliation result — created namespaces, applied tombstones, structured per-scenario info entries, and a removable_entries hint for stale tombstones that are safe to delete from the config. Blocking errors return the structured per-class detail with scenario tags, suggested remediation, and any referencing-script context. wrangler versions upload also forwards exports . Declarative exports lifecycle changes are reconciled when the version is deployed ( wrangler versions deploy or wrangler deploy ), so a versions upload payload can declare new classes in exports without immediately provisioning them. An actor binding ( durable_objects.bindings ) to a class declared only in exports on the same versions upload is rejected with a clear error (code 100406) — the binding cannot be resolved until the namespace is provisioned. Either stage the new class via ctx.exports.X (no binding required) on versions upload and add the binding at deploy time, or use wrangler deploy to provision and bind in one step (the same constraint applies to the migrations flow). Multi-version deploys ( wrangler versions deploy A@50% B@50% ) where the selected versions disagree on declarative exports are rejected server-side with a clear message: deploy the version that changes exports at 100% first, then run the percentage-split deploy. This prevents traffic on one branch routing to code that references unprovisioned or just-deleted DO namespaces. Single-version (100%) deploys are unaffected. Local development ( wrangler dev , vite dev and unstable_startWorker ) reads Durable Object SQLite storage settings from the new exports field, so applications using the declarative flow get correct local-dev storage without needing to also declare a migrations block. @cloudflare/vitest-pool-workers also picks up Durable Object configuration from exports , so tests against an exports -only Worker run with the correct local SQLite storage and can reach unbound Durable Object classes via ctx.exports.X . wrangler types is also aware of exports . Live entries (including expecting-transfer , the receiving side of a two-phase transfer) are added to Cloudflare.GlobalProps.durableNamespaces , which types ctx.exports.X for unbound Durable Objects declared only via exports . #14423 be3f792 Thanks @akshitsinha ! - Add wrangler flagship commands for managing Flagship apps and feature flags. The new wrangler flagship apps and wrangler flagship flags command groups let you create, list, get, inspect, update, set, split, rollout, enable, disable, evaluate, and delete Flagship apps and flags from the CLI, including targeting rules, variations, percentage rollouts, evaluation context, and flag changelogs. #14156 e1532eb Thanks @petebacondarwin ! - Add opt-in OS keychain storage for OAuth credentials By default wrangler stores your OAuth tokens in a plaintext file, and that is unchanged. You can now opt in to encrypting them at rest instead: wrangler login --use-keyring writes the tokens to an AES-256-GCM-encrypted file whose key is held in your OS keyring (macOS Keychain, libsecret on Linux, or Windows Credential Manager). Existing plaintext credentials are migrated automatically on first use. Toggle it with any of: wrangler login --use-keyring / --no-use-keyring wrangler auth keyring enable / disable (or wrangler auth keyring to print the current setting) — useful if you only use named profiles and never run the global wrangler login CLOUDFLARE_AUTH_USE_KEYRING=true|false to override the saved preference for a single command Opting out deletes the encrypted credentials rather than decrypting them back to disk, so you re-authenticate afterwards. The preference applies to every auth profile, and each named profile gets its own encrypted file and key. Per-platform requirements: macOS uses the built-in security tool (nothing to install); Linux uses secret-tool from libsecret-tools (wrangler prints an install hint if it is missing); Windows lazily installs @napi-rs/keyring (~1.9 MB) on first opt-in, and errors with instructions in non-interactive/CI contexts. CLOUDFLARE_API_TOKEN and CLOUDFLARE_API_KEY / CLOUDFLARE_EMAIL continue to take priority over any stored OAuth credentials. Patch Changes #14502 6b0ce98 Thanks @dependabot ! - Update dependencies of "miniflare", "wrangler" The following dependency versions have been updated: Dependency From To workerd 1.20260630.1 1.20260701.1 #14306 bfe48db Thanks @matingathani ! - Remove deprecated --experimental-vm-modules flag and prevent silent exit on unexpected errors wrangler was silently exiting with code 1 on Node.js v26 with no error message shown. This release fixes two independent issues that caused this behaviour: A stale Node.js flag that caused unexpected behaviour on Node.js v26 has been removed. If an error occurs in a situation where the normal error reporting path itself fails, wrangler now always prints the original error to stderr so the cause is visible rather than silently disappearing. #14481 0277bfa Thanks @dario-piotrowicz ! - Improve error message when deploying to a non-existent Pages project in non-interactive mode Previously, running wrangler pages deploy with a --project-name that doesn't exist in a non-interactive context (e.g. CI, piped input) would fail with a generic "project not found" or "This command cannot be run in a non-interactive context" error. Now it provides a specific error message explaining that the project doesn't exist and suggests how to create it. The error also suggests using wrangler deploy to deploy a Worker instead. #14305 98793d8 Thanks @jbwcloudflare ! - Improve asset upload performance with single-file uploads Asset uploads now use a more efficient per-file upload path when the platform enables it. This is rolled out server-side and requires no configuration changes. Existing upload behavior is unchanged when the new path is not enabled. Updated dependencies [ 6b0ce98 ]: miniflare@4.20260701.0
For decades, to be a gamer was to accumulate a lot of stuff. Consoles, controllers, accessories, weird VR gloves that never worked properly, but mostly the games themselves. Over the years, games have come in every shape and size you can imagine. And now that era appears to be ending. On this episode of The […]
Toolbox App 3.6 gives you better control over local storage and makes Windows installation failures easier to diagnose. Clean up removable Toolbox App data from Settings The Toolbox App now shows how much removable data it can safely clean up, providing a breakdown for the download cache, previous versions, temporary leftovers, and leftover tool directories. […]