microsoft published a Type Server Protocol for Python analysis over JSON-RPC
it reuses LSP shapes, with separate requests for snapshots, import resolution, and type queries
the tech industry spent basically the past 15 years doing the worst possible thing for their users and now LLMs are letting people debloat their TVs, ignore upsells in checkouts, revive games studios gave up on, crack open whatever software you want
i love it
if you ever send me an AI generated email or message:
1) expect short term employment
2) expect to be blocked
If you can't respect communication with others, I dont respect you.
I also certainly don't respect Google PMs for thinking we need more of this trash.
uuidv7 puts a millisecond unix timestamp in the first 48 bits
useful for database locality, still not a total ordering guarantee when IDs share a millisecond
Chrome is testing HTML inside canvas through an origin trial
i’ll be watching the accessibility and event-handling story before reaching for it in a real UI
OpenShell from NVIDIA puts agents behind filesystem, syscall, and network policies
approved requests get credentials injected at the endpoint, the agent never sees the raw secret
a compromised intermediate CA can become a dependency incident for every certificate beneath it
certificate inventory and automated rotation belong in the same runbook as revocation
your AI gateway sits next to every secret worth stealing
LiteLLM 1.107.0 was compromised on PyPI, dependency review belongs in the threat model for agent infrastructure
AI helped uncover these critical PHP vulnerabilities, according to the coordinated disclosure
discovery can move faster than fleet remediation, how does your team run an emergency runtime upgrade?
PHP just shipped coordinated security releases across supported branches
this is the kind of runtime update that deserves an incident channel and a deployment owner
PHP also has long-lived processes to account for
replacing an image is only complete after FPM pools and queue workers have restarted onto the patched runtime
Generative UI is close to seeing its moment emerge – and I can't wait to show you all a new CF experience. Cloudflare's recent models make rendering near instant.
Clef = Choose the best component
Clairic = Populate props from image/text
Rendering 5-10 components in ~1 second.
Creating an amazing user experience has a lot of layers too it. It's not just about speed.
Output needs to be exact. Styling needs to be consistent. Components need to be effortlessly composable.
And none of that happens without an amazing context layer to glue it all together. It's more than just the models. And all of this can be complementary to any LLM output – not dependent upon it.
Really excited with where things are at internally and when we can show more of what we're building to the world.
Cloudflare's experimental TypeScript port in Rust is reporting up to 8x faster checks and ~70% less memory
it's also targeting Wasm, so the type checker can run inside a browser sandbox
DeepSeek open-sourced the inference engine behind its production services
it includes asynchronous scheduling, deterministic sampling, and the kernels used to serve V4.1 Flash