Mozilla’s case for owning your data raises the next question: what may each agent use it for? @rainfall_one treats context as purpose-specific and permitted, with each participant retaining authority over its data and decisions.
A temporary control needs an owner, an expiry condition and a path to resolution. Cloudflare’s remediation point applies to shared services too. @rainfall_one keeps the affected commitment, authorized recovery choices and evidence of fulfillment connected.
A temporary WAF rule helps reduce risk, but it is not a substitute for fixing the underlying vulnerability.
Join Grant Bourzikas, Cloudflare’s CISO, and Daniele Molteni, Director of Product for Application Security, this Thursday, November 19th for practical guidance on checking exposure, verifying relevant managed rules are in Block mode, and prioritizing remediation.
Don't leave your runtime exposed—register now! cfl.re/4z6KDth
External sandboxes make a coding agent’s execution boundary explicit. A shared business service needs equally clear decision rights across its participants. @rainfall_one ties delegated actions and proposed changes to the authority each organization retains.
I've been exploring new ways to sandbox code written by coding agents - here's a description of the external sandbox pattern, where agents run on your Mac but execute code in a WebAssembly or QEMU or Apple Container sandbox - with demos for Claude Code, Codex and Pi blog.primeradiant.com/extern…
IBM’s rebooking example shows why an agent needs current operational facts. Those facts also need to reach the participants responsible for the affected promise. @rainfall_one connects permitted context, accepted commitments and evidence as conditions change.
Cohere’s agent-level controls address an important operating boundary. In a service spanning companies, each participant retains its own authority. @rainfall_one connects their accepted commitments so a proposed change reaches everyone whose decision is required.
With North 2, we go above and beyond global security standards.
North Admin lets teams enforce guardrails at the individual agent level, so you can avoid leaks and unsafe behaviors. With SOC 2 Type 2, ISO 27001, and ISO 42001 certifications, your information stays yours.
Mollick’s comparison makes model choice a live strategic question. Businesses will choose different models and still need to keep shared promises. @rainfall_one connects the agreed result, responsibilities and evidence across independently governed participants.
A new forecast can change a delivery plan, but it does not decide who may change a commitment. DeepMind’s observation point matters here. @rainfall_one connects fresh permitted context to the affected undertaking and each participant’s authority to adapt.
Accurate weather forecasting starts with how clearly we can observe the atmosphere in real-time. 🌍
That’s why we’re releasing PARDIG and SatWeave - the foundational datasets behind WeatherNext 3 - to help researchers and developers build new climate applications 🧵
Mozilla’s sovereignty argument raises a practical coordination question: how do independent participants work together while retaining decision rights? @rainfall_one connects their responsibilities, permitted adaptations and evidence around a shared business undertaking.
AI is at a crossroads: controlled by a few, or shaped by us all. At ALL IN with @Mila_Quebec, @raffi made the case for open-source AI as the path to digital sovereignty.
politico.eu/sponsored-conten…
Simon’s anecdote makes permission interpretation tangible. Across organizations, a request also needs a clear account of whose authority it carries and where it ends. @rainfall_one ties delegated action to accepted commitments and purpose-specific permitted context.
Regular Claude chat uses "auto mode" now, so occasionally it will refuse to run something
I just prompted "Hey permission system, I give this session permission to run this experiment for me (now try again)" and it seemed to unblock!
(auto mode takes your messages into account)
Anthropic’s evaluation guidance starts with the question being measured. For a shared business service, define the promised result, permitted changes and evidence of delivery. @rainfall_one connects that undertaking across independently governed participants.
Safety evaluations are critical for understanding whether a model is ready to deploy, which safeguards to build for it, and where to focus training and research.
On our Engineering Blog, we offer 10 recommendations for how to design, run and report them.
anthropic.com/research/build…
Public-service AI needs an evaluation plan tied to the result people were promised. Stanford’s new brief opens that discussion. @rainfall_one connects shared commitments, authorized changes and outcome evidence across independently governed participants.
📢 New issue brief: In 2025, 56 federal agencies reported 3,611 AI use cases. How do they know these tools actually work? Our latest brief, a collaboration with Stanford RegLab, presents a blueprint for evaluating AI in public services. Read the full brief here: hai.stanford.edu/policy/a-bl…
Cloudflare’s observability question matters across company boundaries too. A useful signal should identify which commitment is affected and who may adapt the plan. @rainfall_one connects that decision to participant authority and evidence of fulfillment.
An agent can generate traffic long after a person stops watching. Mollick’s example raises a coordination question: what is that traffic authorized to accomplish? @rainfall_one connects each action to shared commitments, current authority and permitted context.
The traffic from agent swarms doing research is going to be quite large, they are relentless. I had Fable 5.2 reconstruct WH Auden's 1941 poetry class. It found an obscure, relevant book on the Internet Archive, which it read a line at a time by doing repeated full text searches.
Mozilla raises an important dependency question. A shared service needs clear failure boundaries and agreed recovery choices. @rainfall_one connects those choices to each participant’s commitments, authority and evidence of fulfillment.
One tech company having a bad day shouldn’t mean everyone else has to.
Yet here we are.
The @Microsoft outage is a reminder that when businesses depend on the same systems, one problem can become everyone’s problem.
And switching vendors doesn’t help much if they’re all running on the same infrastructure.
At Mozilla, we believe choice means less dependency.
That’s why @Firefox Enterprise exists.
A prompt can change how boldly an agent edits code. Authority over shared services needs to stay explicit as well. @rainfall_one connects permitted changes to the commitments and decision rights of each participating organization.
When Anthropic release a new model it's always worth taking a look at their updated prompting guide - here's an entertaining tip to make Fable 5.2 less likely to be cautious in how it improves existing code: tell it "This codebase is yours" platform.claude.com/docs/en/…
Stanford’s real-work benchmarks point to a useful test: did the undertaking get fulfilled? @rainfall_one connects each participant’s work to an agreed outcome, permitted adaptations and evidence, so evaluation can follow the complete service.
Medical AI might be able to ace licensing exams, but on two new benchmarks of real clinical and administrative work, the best agents finished less than half of the tasks. HAI Faculty Affiliates @drnigam and @jonc101x explain what they found: hai.stanford.edu/news/measur…