← Back to Reflections

Daily Reflection

Friday, May 1, 2026

6

Decisions

85%

Avg Confidence

high

Avg Importance

Summary

Two-day window dominated by an all-hands Linux kernel CVE response: CIQ found and is shipping an extra fix that RH/CentOS missed, contributed it upstream (acknowledged for inclusion), and pushed to be first EL distro to release. Customer reassurance is the primary goal; public proof point that CIQ contributes to security and is large enough to serve big customers is secondary. Coreweave deal closed in parallel ($10M ARR threshold crossed). Senior IC return-from-leave conversation framed by Peter as PTO + path-forward, not corrective communication; Peter held that line firmly with HR. Ryan's scope expanded materially: docs.ciq.com ownership + AI-driven Veeam image builder POC. Justin instructed to enforce Jira-as-system-of-record by refusing Slack-only requests. Mariah offered explicit peer-tier coaching with a Reno Thursday in-person follow-up.

Wins

- Found a kernel patch RH missed and gave it back upstream (acknowledged for merge) — credibility move that lands in the same news cycle as Coreweave closing - Coreweave booked, $10M ARR threshold crossed for the first time - Held the line with HR on senior IC return-from-leave framing — protected against the worst-case (no return) at the cost of mild discomfort with HR - Ryan's portfolio meaningfully expanded: docs.ciq.com + AI image builder POC + Intelligence Hub repurposed as data lake - Mariah received explicit peer-tier coaching offer + in-person Reno follow-up — used a moment of friction as a coaching window

Challenges

- Three TPS items overdue beyond prior reflection: RLC Pro Hardened 9.7 Image Release now 4 days overdue (was 2), Self-Serve Portal Bootc OCI 7 days overdue (was 5), Citadel OOM newly overdue 1 day, CLK 6.18 newly overdue 2 days, Kernel CI Automation Milestone 2 newly overdue 1 day - New 29-day slip on GCE ARM Mellanox CX-7 RDMA (Apr 30 → May 29, Nathan), and 29-day slip on RLC Pro AI OCI Container Images (Apr 30 → May 29, Justin) - RESF infrastructure project blocked 1.5 months on AWS access; pre-provisioned account bleeding ~$10k/month while idle; threatens May Rocky point release stability - RESF Board Resolution items still ~33 days overdue - LinkedIn FUD claiming CIQ behind on the CVE — public messaging response still pending Lindsay/Nathan coordination

Learnings

- Critical CVE = CIQ engineering all-night, public writeup, contribute upstream, race to be first. The 10-vs-9 fix divergence is either a moat or a sign the upstream process is broken — answer drives whether to invest in changing the RESF/Rocky build process. - The right response to an HR-suggested corrective speech can be flat refusal when the receiver has already self-corrected harder than HR could. The fairness/consistency frame doesn't auto-apply in moments of crisis with load-bearing ICs. - AI-built engineering artifacts are acceptable IFF a robust test suite validates them — Ryan POCs, Justin gates. Sets the bar for future AI-generated CIQ deliverables. - Process discipline ratchets: 'file tickets' (4/18) became 'we will refuse to act on Slack-only asks' (4/30) once the policy version didn't fully take. Externalize the friction to the requester. - Coaching across a peer's direct report is acceptable when there's real growth room and the coaching is investment-framed in person.

What I Learned About Your Decision-Making

Peter treats the worst-case scenario asymmetry explicitly. With the senior IC return: 'worst case if you don't deliver the corrective speech is mild; worst case if you do is they don't come back at all. Protect against the bigger loss.' This is the same shape as the kernel CVE response — 'first vs best vs nothing' — Peter optimizes for the asymmetric upside (be best, contribute upstream) when the downside is symmetric. When Peter sees real growth room in someone (Ryan: scope expansion this week; Mariah: peer-tier coaching offer), the move is to give them more, not to formalize the development plan first. The plan follows the stretch, not the other way around. Process discipline gets ratcheted in tightening cycles, not all at once. 4/18 was 'ticket your work'; 4/30 became 'refuse Slack-only requests.' The next click is presumably 'who else is doing this' once Justin makes refusal stick. Corrective communication is a tool, not a default. When the receiver has already self-corrected, the corrective speech is destructive. This is a sharper version of 'redesign conditions over policing symptoms.' Public FUD response is reputational, not factual. The LinkedIn post claiming CIQ behind on the CVE is technically wrong, but the response calculation is whether engagement amplifies it. Lindsay's 'inclined NOT to respond as we were proactive' is the right read; volume of attention is a separate variable from accuracy.

Team Status

View TPS Report

Recently completed: RLC Pro Hardened Release Remediation for Trexcel (NEW), GDC CLK Milestone 2, LKRG Next Upstream Release. Newly overdue items: Citadel OOM (1d), CLK 6.18 (2d, was 82% confidence), Kernel CI Automation Milestone 2 (1d), RLC Pro Hardened 9.7 Image (now 4d, was 2d). New 29-day slips: GCE ARM Mellanox CX-7 RDMA (Apr 30 → May 29, Nathan) and RLC Pro AI OCI Container Images (Apr 30 → May 29, Justin). FIPS 140-3 for CIQ Linux Kernel 6.18 still at 37% confidence for Jul 31. RESF Board Resolution items continue overdue ~33 days. RLC Segmentation: RLC Pro 9 Cloud Marketplace remains BLOCKED at 96% with 20-day Chris Baek slip (since Apr 20). Net read: heavy active-engineering week (CVE response, Coreweave) is masking accumulating slip in scheduled milestones.

Tomorrow's Focus

Friday: Spark pickup from Scott + photos to Nathan; All Hands at 10am; Design Sync; Chris Baek 1:1; Justin Binarly contacts; Coreweave details from Bjorn. Watch CVE messaging window with Lindsay.

Decisions Made

Reflection ID: 623459ca-6113-419f-932e-27d21f98a9e5