I take over high-stakes technology programs. Especially when legacy data, internal politics, and regulatory pressure have outgrown the original plan. AI is one tool; delivery is the mandate.
01
Large technology programs rarely fail because the strategy deck is weak. They fail when legacy data, vendor promises, internal politics, and regulatory exposure meet a deadline the organization can no longer move.
That is where I operate: after the easy answers are gone, before failure becomes the accepted outcome.
I bring Silicon Valley execution into traditional, high-consequence environments. Startup meets enterprise, without external consultants.
I take responsibility for the constraint, make the data usable, narrow the scope to the smallest live slice, and harden the system until executives, operators, and control functions can rely on it. AI matters when it helps the work ship.
02
Start with data quality, availability, permissions, and the rulebook before choosing a model.
Put the smallest useful system into production fast enough for reality to test it.
Make it robust enough for users, audit, compliance, and the next organizational change.
03
Production-grade AI where data is messy and compliance is non-negotiable — without consultants.
04
I keep operating notes on problems worth solving: AI in constrained environments, turnarounds, and the discipline of making complex systems simple.
No weekly fluff and no recycled commentary. I send a note only when there is a concrete build, decision, or operating lesson worth an executive's time.
Leave your email to receive the first issue. No spam, just signal.
05
What's on the operator's desk this cycle: two articles, one Reddit thread, one shipped repository per hub. Pipeline runs daily across a hand-picked source set; drafts are written by Claude (Anthropic API) against an operator-voice prompt and auto-published. Staging gates production: a bad day on staging keeps yesterday's view live.
Sequoia thesis: agents that close the loop, not chat windows
Capital is shifting from generic chat experiences to agents wired into billing, CRM, and ERP. Sign-off authority is now the bottleneck — not model capability.
Operator implication. European banks and insurers should expect agentic-ERP procurement asks within 12 months. Build the sign-off workflow before the agent.
Read article →
EU Enterprises Are Running Contingency Drills on US Hyperscaler Dependency — With No Clean Exit
A practitioner thread from EU-based DevOps engineers surfacing how companies in regulated sectors are actually handling AWS/Azure/GCP lock-in risk today — not in theory. The discussion reveals that most EU enterprises have no credible Plan B: switching costs, skill gaps, and the absence of competitive European cloud infrastructure keep them exposed. This is the operational reality behind the political noise about digital sovereignty.
Operator implication. If your enterprise has not formally stress-tested what a forced reduction in US hyperscaler dependency would cost — in migration time, reskilling, and contract penalties — you are carrying a regulatory and geopolitical risk that belongs on your board agenda now, not after a transatlantic policy shock.
Open thread →
Microsoft's May Patch Tuesday: 30 Critical CVEs, No Zero-Days — But the Patch Window Is the Exposure
Microsoft's May 2026 Patch Tuesday dropped 30 critical CVEs — the largest critical count in recent cycles — with no confirmed zero-days actively exploited at release time. The risk window is the gap between release and enterprise deployment: attackers reverse-engineer patches within 24–72 hours to build exploits targeting unpatched systems. For regulated European enterprises running legacy Windows estates with extended change-approval cycles, this Patch Tuesday materially widens the attack surface this week.
Operator implication. If your patch approval cycle runs longer than 72 hours, triage the 30 critical CVEs today and push emergency change requests for any that touch RCE or privilege escalation vectors — your regulators will ask about patch lag if an incident follows.
Read article →
Google Cloud API Billing Gone Rogue: Unauthorized Usage Charges Hitting Customers with No Clear Remedy
Multiple Google Cloud customers are reporting surging API bills from usage they did not authorize, with Google's dispute and refund process producing little resolution. The pattern — leaked or scraped API keys triggering runaway consumption charges — is not new, but Google's apparent inability to provide timely remediation is the signal here. For enterprises with Google Cloud API keys distributed across dev teams or third-party integrations, this is a live financial control failure, not a theoretical risk.
Operator implication. Audit your Google Cloud API key inventory now — enforce per-key spend caps, rotate any key exposed in CI/CD pipelines or third-party tools, and confirm your cloud contracts define the liability boundary for unauthorized usage charges before your CFO sees an anomalous invoice.
Read article →
06
// 06 · discuss with the operator
One short turn. The bot speaks in the same operator voice as the signals above and is grounded in this week's Pulse. Three free turns per visitor; after that, drop your question to Felix directly.
07
Reach me directly for a strictly confidential exchange about your current constraints. Whether it's a failing delivery or a critical data mandate - let's evaluate if I can take it over.