New post: I added Direct I/O support to Cassandra for compaction reads, cutting p99 read latency by 5x (and p50 a little).
Walks through the page cache mechanics, benchmarking, and when to enable it. Some funky charts along the way.
https://t.co/wvsVSGJVj7
There is no silver bullet or perfect prompt but if you do hard core engineering (not related to LLMs themselves but instead how you apply them) the craft gets recursive very very fast.
People keep asking me about what skills I use and how I prompt LLMs and the answer shocks folks.
I don’t pre-install any skills or rely on attempting to eval max markdown into a context window. There’s no magic trick except application of the craft of software engineering itself!
Stop chasing LLM prompt fairies and instead focus on the craft.
Our paper, "How to Write to SSD", received a Best Research Paper Honorable Mention Award at VLDB ’26!
Bohyun (Lia) from TUM did a phenomenal job on this work. Check it out if you're interested in modern NVMe SSD internals and how to get the most performance out of them.
@British_Airways our flight was cancelled 2 hours before departure and your phone lines have been busy since, with no choice to wait in a queue. How do we speak with someone?
@British_Airways We have rebooked with another airline as the only option was tomorrow. How can I contact reservations to indicate we still want to return, and also be told the cancellation reason?
If you're interested in SSD internals and how to use them efficiently, our paper, “How to Write to SSDs,” has been accepted to VLDB and is currently on the Hacker News front page. https://t.co/bwYAtPFTBg
During my time at @awscloud I tried to be an astute observer of how the Senior Principal and Distinguished Engineers carried themselves in new projects/teams and design reviews. This has made a lasting impact on how I operate as well as evaluate others.
All of these "Very Senior" engineers are smart but the absolute best ones were humble, secure, incredibly curious, and appreciated the history and path of a system. They can jump into whole new domains and immediately win over the people under them that ultimately do the work.
The worst Very Senior engineers show up and inflict their insecurity by asking divisive and superficial questions like "why didn't you use {{language foo}} or {{database bar}} or {{platform qux}}?" These very questions reveal their inner fear: they don't have the depth and breadth their position requires.
As a manager, it took me far too long to learn that meetings aren't actual work. I thought meetings drove output. Turns out, most meetings are about control, not collaboration. About signaling ambition versus pursuing it. If you can't pass these three tests, cancel the meeting: