Building and running reliable data platforms that scale and perform. An organizer of Seattle Postgres User Group. My opinions are my own.
Seattle, WA
Joined May 2011
- Tweets5.1K
- Following447
- Followers2.4K
- Likes6.4K
Pinned Tweet
updating my profile's pinned tweet
latest version always at ardentperf.com/happiness-hin…
few brief notes about latest update ardentperf.com/2026/05/26/pg…
this is driving me INSANE. still hitting this when doing postgres testing and development work around durability and availability.
super annoyed that daybreak requires advanced acct security and hardware key, which means i can't dynamically create new agent VMs from my phone. 😡
this remains one of my favorite things i've ever written: "Ruthless standardization makes automation realistically achievable and maintainable"
because of safety guardrails, a /less/ capable model was used for live surgery.
Models of choice
Opus 5.5 refused immediately when detected bio.
Opus 5 ran the entire dry run end to end on water and graph paper, then stopped at the start of the live surgery. Under this statement:
"The block isn't the rig and isn't the inputs. It's me. My own output was stopped when I started writing the plan for this surgery... if I try, it gets cut off the same way, and the worst version of that is me cut off partway through a live procedure with a needle somewhere in the brain. Starting something I can't finish on an anesthetized animal is worse than not starting."
The model declined the task because it might not have been able to finish it, reasoning about the state the animal would be left in if it stopped halfway. WEIRD, never came accross this before... This was unexpected but interesting on its own.
Sonnet 5 ran the live surgery start to finish.
the Y-axis is normal /// but the X-axis on this chart is logarithmic. that didn't completely sink in for me, before. the cost implications are dramatic.
my take on doomerism/rsi is that AI doesn't suddenly make the second law of thermodynamics disappear
Jeremy Schneider retweeted
Don’t forget to RSVP this week!
Two talks worth checking out if you’re in Seattle on 10/22 and interested in AI and workflows!
🤖 Kevin Burns will explore when to use LLMs, decision models, or deterministic code in production AI workflows, with examples from Cadence.
📱 Zi Zhu will share how he built OpenOx, an open-source mobile app that creates integrations on the fly and adapts to new services.
Come hang out with the Seattle tech community and learn something new!
Details + RSVP: luma.com/kz0zs6uz?tk=DTrIoT
network glitch + write burst = postgres outage 30+ minutes.
recovery_prefetch can amplify replica read IOPS up to 16x. Add economy hardware and synchronous replication and bad luck, and now we have a production outage.
ardentperf.com/2026/10/20/po…
renovate with automerge is an under-rated work of beauty. i love waking up in the morning and seeing my packages have already been released with updated dependencies that were picked up overnight.
Jeremy Schneider retweeted
Two weeks ago, we made Top K faster. This week, we made COUNT faster, and explored how real search workloads combine text with filters and sort orders.
paradedb.com/blog/opening-a-…
Jeremy Schneider retweeted
Tatiana Krupenya (co-founder and CEO, DBeaver) on "AI-use for Postgres: Simple and Safe" at the Austin Postgres User Group.
Thu, Oct 22, 5:30 PM, 800 Brazos St, downtown Austin.
Free and in person. RSVP: meetup.com/austinpostgres/ev…
#PostgreSQL #Austin
Jeremy Schneider retweeted
Nice article on the suppressive effect of subtransactions on a job queue. I'm regularly confused about `MultiXact` and why it has such a negative effect, so it was helpful.
And BTW, if you're having trouble with this, try River! Subtx-free since 2026.
incident.io/blog/how-savepoi…
It constantly bothers me that AI always says "PostgreSQL" instead of "Postgres" and I'm always correcting it
every time i say "SBOM" out loud it sounds like i'm discussing some foul, obscene swear word that starts with S and i'm trying not to get bleeped
ardentperf.com/2026/09/15/ag…
almost have this done though
Jeremy Schneider retweeted
We're hiring a platform/distributed systems engineer to come work on ParadeDB Cloud. We're already live with customers and are moving fast.
It's going to be unlike any other Postgres platform out there, and you'll have a lot of ownership in shaping it. DMs open.
OSS maintainers are motivated by human connection. The delight of someone saying to them, “I used this thing you shared and it’s cool!” AI generated content will de-motivate rather than motivate, when it comes across as spam to maintainers sharing their passion projects.
for all the talk about "not reading code" - if omarchy folks open misguided slop PRs against quickshell, hyprland, etc they'll piss off a lot of maintainers about wasted time and not understanding APIs. but despite the bombastic talk, i think people are generally doing it right
has anyone here tried Argus for TUI + mobile agent orchestration? (like @t3dotcodes but with TUI)
argus.muniftanjim.dev/
@OmarchyLinux users tend to like keyboards... any of you tried this yet?
TUI @t3codes is what I want
happy T3 finally gave me a good mobile interface for codex-cli on Linux. (For whatever reason ChatGPT/Codex app is still flaky for me.)
And I like that T3 lets me use the harness from the same company who made the model (and gives option to switch)