Leaving Intel

Reactions to the Departure & Reputation

• Many commenters see the author as one of the top performance engineers and expressed that Intel is losing a major asset.
• Others found the farewell post underwhelming or overly self-promotional, arguing the accomplishments list didn’t “move the needle” for them.
• Several people countered the criticism, pointing to the author’s long-standing contributions (books, tools, techniques) and calling him an outlier in productivity.

Intel’s Health and Talent Drain

• Multiple comments frame this as part of a broader trend of strong people leaving Intel, with some comparing Intel’s trajectory to IBM, Kodak, or other “slowly dying” giants.
• Others push back, noting Intel’s large market share and ongoing silicon sales, plus government and strategic backing, and argue Intel is “not going anywhere,” even if it has serious competitive problems.

Compensation, Status, and Where He Might Go

• There’s speculation he’ll join a major AI or GPU-focused company; some believe his contributions could easily justify extremely high compensation.
• Debate arises over whether Intel Fellow–level compensation is low relative to FAANG “fellow” tiers, and how much exceptional engineers should earn.

Self-Promotion, Metrics, and “Impact”

• The meticulous counting of meetings, initiatives, and recommendations triggers a large thread on score-keeping culture at big tech.
• Some see this as necessary for remote and senior roles to prove impact; others view it as hollow metrics and “Byzantine bureaucracy.”
• There’s a broader argument over how to fairly measure impact, the limits of metrics, and the politics of performance reviews and promotions.

Remote Work, Time Zones, and Meetings

• Prior posts about 1–6am meetings are cited; commenters highlight the real toll of being extremely remote from HQ.
• Some sympathize; others are surprised someone with his reputation still had to “prove” remote effectiveness.

Technical Tools and AI/GPU Focus

• GPU flamegraphs for AI workloads are generally praised as an obvious and useful extension of earlier work.
• One subthread debates how novel and essential flamegraphs really are, and whether they are oversold versus the underlying kernel profiling.

Side Tangents: Retro Gear, LLMs, and Cloud Performance

• The office photo spawns a nostalgic digression on Commodore/VZ retro hardware.
• Another subthread discusses using LLMs to explain 6502/Z80 assembly: helpful with enough system context, but often hallucination-prone.
• A separate technical tangent explores why virtualized Intel instances on AWS can be ~20–25% slower than bare metal, with suggestions to use profiling and top‑down microarchitectural analysis.