Engineering Notes & Build Logs
Articles and build-in-public notes on systems integration, embedded, robotics, automation, and digital twins. By Andrew Michelis at KnackMentor.
In Cyber-Physical Systems, Latency Is a Correctness Property
A control loop can be rock solid in simulation and unstable on hardware over a few milliseconds. Why timing is part of whether the system is right, not a performance detail to tune later.
Teaching a Vehicle to Distrust Its Own Sensors
An integrity layer built in the open: how a vehicle notices a lying sensor, decides when to stop, catches the quiet faults the fast check misses, and proves it honestly, with the tuned core behind one seam.
A Small Language for the Agreements Between Systems
HashTag is a tiny open notation for the agreements between parts of a system, so a script can check an assembly holds together before you build it.
How I Made a Vehicle Notice Its GPS Was Lying
The M1 milestone of the open autonomy build: a simple statistical check that catches a spoofed GPS at the first bad reading and keeps it out of the estimate.
Save the Web Exactly as You See It
Print-to-PDF re-renders the page and loses what you saw. WebClip, a free open-source Chrome extension, saves any page to a pixel-faithful PDF, all on your device.
Building a Sensor-Fusion Testbed You Can Trust
The first build increment of an open autonomy project: fusing GPS, IMU, and wheel speed into one estimate, and proving it tracks the truth before adding any fault detection.
What I'm Building in the Open, and Why
The plan for an open, simulation-first project on a hard problem in autonomy: how a system knows when it can't trust its own sensors, and fails safe.
What the Hardware-Software Boundary Taught Me About Risk
Fifteen years where hardware meets software taught me that the biggest delivery risks hide in the seams between disciplines, not inside them.
The Trust Ladder: Why Diagnostic-First Beats Sales-First
Why diagnostic-first commercial motion creates stronger fit, better decisions, and less wasted effort for technical teams.
Stakeholder Alignment Is Not a Meeting: It Is Evidence
Why alignment claims collapse unless the relevant decision, ownership, and readiness evidence are visible.
What a Bounded Diagnostic Review Actually Delivers
What a diagnostic-first review does, what it does not do, and why that boundary matters.
When Technical Debt Becomes Delivery Risk
The moment technical debt stops being an engineering nuisance and starts threatening releases, milestones, and credibility.
The 3-Minute Diagnostic That Surfaces Your Delivery Blind Spots
A short diagnostic lens for surfacing the risk clusters teams usually notice too late.
Validation Gaps: Why “We Tested It” Doesn’t Mean “It Works”
Passing tests aren't proof of readiness. The structural difference between testing activity and validation evidence, and where the gaps usually hide.
How to Know If Your Team Is Actually Ready for the Next Milestone
Most teams that miss a milestone were "ready" the week before. The gap between activity and evidence is where readiness fails. Here's what to check.
Hidden Assumptions That Derail Software Delivery
The assumptions teams leave implicit, and why they only become visible when the schedule is already under pressure.
Interface Ambiguity: The Silent Killer of Technical Programs
Why teams keep moving while shared boundaries stay undefined — until integration forces the disagreement into view.
The Readiness Question Nobody Asks Before a Release
The one question that exposes whether a team is actually ready or just close to shipping.
5 Delivery Risks Hiding in Your Engineering Team Right Now
Five delivery risks that hide inside engineering teams until integration day. What to check before they surface as rework.
Why Integration Milestones Fail — What to Check
Status can look green while readiness is still weak. This article shows the patterns that break milestones late and what to check earlier.
The Bug That Wasn't in the Code: A Sensor-Integration Story
The code was correct and the system still failed. A generalized, anonymized story about where integration bugs actually live, and why they hide from code review.
Sensor Fusion at the Systems Level, Past the Tutorials
Filter tutorials stop where the hard part starts. What sensor fusion actually demands once it has to hold up across real sensors, timing, and faults in a live system.
Building in the open
A new build log lands as each milestone ships: sensor fusion you can trust and catching a spoofed GPS (both done), then a fault taxonomy and a safe-state layer, a dynamics twin that catches the slow drifts, and a robustness evaluation.
Connect on LinkedIn →