excel user & data enthusiast. living the dream at MotherDuck
Joined July 2008
- Tweets 31,743
- Following 2,017
- Followers 8,434
- Likes 100,521
3,239 Photos and videos
14h
Gotta shoot from behind half court so you donβt get that negative fg%
1
298
23h
The people love low stakes gambling
is everyone just claude maxxing or what
3
499
Jun 12
always be eval-ing. or something.
Anyway i built a dashboard (using a MotherDuck dive, of course) to explore my evals and keep things on the rails.
Good news: it worked - i was able to get my eval-maxxing to work.
1
4
556
Jun 12
if this type of stuff interests you check out the interactive version of it
motherduck.com/dive-gallery/β¦
298
Jun 12
did some looking into which players are winning the most on daily82.com
discovered that i need to make some tweaks
3
1
1,217
Jun 12
so the plan is we make a test harness and tune parameters a bit, will drop the changes soon.
1
261
Jacob Matson retweeted
Jun 12
One prompt. Two production pipelines and a dashboard.
Mehdi Ouazza put Flights through its paces: he gave Claude a single prompt and let it build his entire data stack on MotherDuck.
π¦ A Postgres replication Flight that moved 60M rows, with auto-retry
π¦ A Hacker News API Flight that runs in about 6 seconds
π¦ A Dive dashboard to finish
The agent writes the Python, deploys it, schedules it, and verifies its own work along the way. MotherDuck handles the runtime, secrets, and scheduling.
3
17
1,216
Jun 12
Not my proudest work butβ¦
82-0 π Daily 64-18 β same five rolls, beat my record: daily82.com/d/2026-06-12?s=Wβ¦
213
Jacob Matson retweeted
Jun 11
What Iβve learned from playing a hundred hours of 82-0
- Moses Malone was a beast
- Isaiah Thomas (DET)β¦not amazing stats for your lore
- How did Steve Nash only score 17 ppg and 10 agp in a decade with 2 MVPs?
- George Gavin mustβve been a walking bucket
- Bernard King was the best player for like 3 different franchises
- Charles Barkley, sorry for not respecting your game
- Indiana has zero players in their franchise history with amazing stats
238
265
8,486
705,951
Jun 11
This is a really interesting take, but I disagree for one key reason: intelligence of these compounds. In a competitive environment, you will fall behind if you are using a tool that is 80-85% as good. If you are in a niche / have a monopoly, thatβs fine.
Jun 11
So you can use the 5th/6th/7th best LLMs, getting 80-85% of the top guys' performance, but at an 85-95% discount in price?
You know what we call that? A commodity...
exactly what happened with LCD TVs, OLEDs, solar panels, electric cars, phones, etc
good luck with your AI IPOs!
4
734
Jacob Matson retweeted
Jun 11
Frog put the model in a box.
"There,β he said.
"Now it will not answer about cybersecurity, biology, and AI research."
"But you still can open the box," said Toad.
"That is true," said Frog.
1
3
317
Jun 10
working on making things faster, its a good exercise
this is the first thing i'm thinking about throwing fable at - optimization problems like this.
1
430
Jacob Matson retweeted
Jun 10
β―β² Introducing Flights, a new agent-native data pipeline tool
Author in python / SQL, trigger in SQL MCP on a schedule. I use it for data pipelines, but also for agent workflows, classical ML tasks, screenshotting Dives and posting to Slack, you name it.
Give it a spin!
1
6
46
2,883