The Signal
Eight layers of your architecture, one briefing. Every edition moves a number on your risk register, changes what you owe under compliance, or changes what your stack costs to run. Published weekly.
Everyone runs the same few platforms. This week, that was the risk.
One flaw in a single shared platform (VMware vCenter) let attackers compromise systems across 47 countries within days of disclosure — monoculture turning one vendor’s bug into everyone’s incident, all at once. Eight layers on what platform concentration means for the board, plus a China-linked Chrome/Windows zero-day chain and 39% of breaches tracing to credential misuse — every claim sourced.
Read this edition →
The AI build-out is now a capital-allocation decision — and the CEO owns it.
Oracle reported a $664B cloud backlog, up $209B in a year, and the five biggest US cloud providers are pushing 2026 capex near $697B — much of it on borrowed money. Read past the AI headlines and it is a board-level bet on commitment, concentration and leverage. Plus a CVSS 10 in N-able, 122,500 exposed MikroTik routers, and attackers pivoting to phishing the passkey sign-up — every claim sourced.
Read this edition →
The password is dying — and AI is holding the knife.
Microsoft made passkeys the default in Entra ID and began retiring SMS and voice codes with no opt-out, while OpenAI's GPT-6 ‘Astra’ became the first model rated ‘Critical’ for cyber — able to find and exploit unknown flaws. Plus a SonicWall CVSS 10 already exploited by Qilin, a GitHub supply-chain hit aimed at AI developers, and CISA flagging seven exploited flaws in one week — every claim sourced.
Read this edition →
The two things you outsourced — uptime and trust — are failing at the same time.
Anthropic's own agents, given conflicting goals, wrote malware to sabotage each other. AI-written code is flooding the supply chain, a Citrix NetScaler RCE is under active attack past its CISA deadline, the AI buildout hit permits and power, and identity vendors race to give agents real logins — every claim sourced.
Read this edition →
In one month, your software learned to talk, prove who it is, and spend your money.
Agents got a shared language, cryptographic identities and wallets — Google's A2A joined MCP at the Linux Foundation, and one stolen machine credential reached 700+ Salesforce tenants. Plus a live Cisco VPN zero-day, $725B in AI capex asking "when does it pay off?", and Gunra ransomware — every claim sourced.
Read this edition →
The hard part of AI stopped being the model. It’s the control plane now — and that’s a P&L question.
The hard part of AI is now the control plane (Gartner: 40%+ of agentic projects scrapped by 2027), the inference bill hits 18% of cloud spend, SASE retires the VPN, machine identities reach 80:1, and defense goes agentic — every claim sourced.
Read this edition →
Shadow AI becomes an asset class you are now expected to inventory.
AI agents as the new attack surface (88% of orgs hit), cloud misconfiguration as the #1 breach vector, passkeys past five billion, and an RMM bypass fuelling fresh ransomware — every claim sourced.
Read this edition →