james (@james_d) header
james (@james_d) profile picture

Engineer @x

New York, USA
Joined June 2022
Photo by james (@james_d)Photo by james (@james_d)Photo by james (@james_d)Photo by james (@james_d)Photo by james (@james_d)Photo by james (@james_d)Photo by james (@james_d)Photo by james (@james_d)Photo by james (@james_d)Photo by james (@james_d)
Pinned Tweet
just setting up my 𝕏 Money nitter.cf/i/jf/cards/money/share…
12
2
4
53
5,428
My agents may be able to code better than me but it’s beautiful weather in Washington Square park today
4
14
645
james retweeted
Replying to @RobertJBye
good engineers are good designers, good designers are good engineers
1
1
25
926
look mom i'm famous
Flex your cashback on the timeline Go to Activity, select a transaction, then tap Share to X
1
1
27
1,395
Proof of coffee. Thanks guys 🫡 You can explain to my coworkers why I’m shaking at my desk
1
6
837
Are you still babysitting your agents? Then you can't just jump straight to fully agentic loops. Things will break. You have to build to up to it step by step.
11
577
You don’t need to keep the same AI chat alive for months. Have your agents write project state to shared files, then tell new chats to read those files before getting started.
1
12
611
A lot of AI coding setups have a big bottleneck: the human in the loop. Even if you become 10x engineer, you're still limited by your time. So what if you took yourself out of the loop and ran your agents like a team? Recorded on my coffee run through the Village this morning
8
1
28
2,551
@allegrajacchia told me to try vertical videos. First one, let me know how I did 🙏
2
4
383
So I misread the label on my vitamin D and have accidentally been taking 25,000 IU of vitamin D every day this week. Oops. Apparently 125mcg is 5000 IU not 1250. They should have taught me that at Waterloo instead of making me compute Lagrangian duals.
4
9
678
The rest of my team, when I show them the latest UI I've built to manage my agents
2
4
546
I'm only talking to my agents in the lambda calculus
1
8
456
"The bitter lesson is based on the historical observations that 1) AI researchers have often tried to build knowledge into their agents 2) this always helps in the short term, and is personally satisfying to the researcher 3) in the long run it plateaus and even inhibits further progress 4) breakthrough progress eventually arrives by an opposing approach based on scaling computation by search and learning. The eventual success is tinged with bitterness, and often incompletely digested, because it is success over a favored, human-centric approach."
3
9
424
I’ve always had a lot of respect for martini drinkers. That’s someone who knows where they are going and wants to get there as quickly as possible. I need to feel the heat from the gin martini in my bloodstream by the time the appetizer hits the table
26
30
6
1,422
94,348