Midwestern fella teaching folks how to code. Education, code and philosophical pragmatism. Past: @MissionBit @DevBootcamp @Everlane https://teapawt.net/t.co/NInB7ebHrd
San Francisco, CA
Joined May 2007
- Tweets12.6K
- Following1.1K
- Followers3.5K
- Likes14K
Pinned Tweet
Programming is hard *and* anyone can do it. Telling beginners that it's easy makes them feel stupid when they struggle.
Worthwhile things are often hard: writing well, playing an instrument, cooking, etc. "All" they require is the right kind of practice and enough time.
/jesse-farmer.php?q=๐ retweeted
Here are some opinions:
a) direct instruction is a good way to think about teaching
b) if I could use my judgement, I wouldn't mind being handed a teaching script
c) if you want to see how a script can go wrong, look at the "high quality instructional materials" fad in the US
I've seen lots of people talking about SHEIN's acquisition of Everlane over the last week. Nobody I know was happy about it.
But the story isn't over. Yesterday, Michael announced stillradical.com/ which @KimBhasin covered in the @nytimes
nytimes.com/2026/05/28/businโฆ
/jesse-farmer.php?q=๐ retweeted
We had an unbelievable tech team and talent overall. Many of them went onto awesome careers.
Founders of Coderpad, Modern Treasury, Maude, amongst others
EPD at Airbnb, Stripe, Apple, Oscar, Google, Iโm sure I am missing some.
Many were fist-time PMs and engineers and Iโm very grateful to have helped them get started.
/jesse-farmer.php?q=๐ retweeted
S/O Everlane one time.
2013-2016.
You just had to be there.
Best first job I could ever ask for.
Claudeๅ
็ใฎๅฟๆใชใ็็ด็กๆฏใฎใ้ซ่ชฌใๆๆฆ่ณๆฅตใซๅญใๅฅใใพใใ๐โโ๏ธ
I've never seen Claude do this before! It's the third question in a series of questions about Cartesian trees.
This is Opus 4.7 set to "Adaptive". Does no visible thinking block imply no extended thinking?
ALT A screenshot of Claude answering a question given in English using Japanese. > User: > Any relationship between Cartesian trees and union find? > > Claude: > ็ดๆฅ็ใช้ขไฟใฏ่ใใใใฉใ่ๅณๆทฑใๆฅ็ถใใใใคใใใใ > Sorry โ let me restart in English. > There's no deep structural identity between Cartesian trees and union-find, but they intersect in a few specific algorithmic places worth knowing.
/jesse-farmer.php?q=๐ retweeted
Itโs really sad, the average vizier only knows about 34% of the total secret passageways in any palace. If youโre onboarding into a new sultanโs retinue I would recommend as a first course of action the comprehensive mapping of as many subterranean tunnel systems as you can
/jesse-farmer.php?q=๐ retweeted
This was a tough but necessary decision - I posted my own notes on this here, from the perspective of a current PSF board member simonwillison.net/2025/Oct/2โฆ
TLDR; The PSF has made the decision to put our community and our shared diversity, equity, and inclusion values ahead of seeking $1.5M in new revenue. Please read and share. pyfound.blogspot.com/2025/10โฆ
๐งต
pyfound.blogspot.com/2025/10โฆ
/jesse-farmer.php?q=๐ retweeted
This is empirically incorrect.
Hundreds of thousands of fast.ai students have learned the required math for ML as they go.
By *far* the biggest problem we've seen is from people who try to learn the math first. They learn the wrong stuff & have not context.
I feel like direct instruction advocates often have a case of "grade school"-itis.
My experience teaching adults, for example, is that they have a wide range of prior experiences to draw from. Some use math to learn ML, some use ML to learn math, and some bounce between.
This is empirically incorrect.
Hundreds of thousands of fast.ai students have learned the required math for ML as they go.
By *far* the biggest problem we've seen is from people who try to learn the math first. They learn the wrong stuff & have not context.
/jesse-farmer.php?q=๐ retweeted
Itโs insane whatโs going on at the University of Chicago, and this is well worth reading. compactmag.com/article/the-cโฆ
I love automaticity and agree it's powerful. I also have a math degree and still spend 2-3 hours/week working through some math textbook or paper.
If you ask me to calculate 8*12 then I'm doing "8*10 + 8*2". Every time.
So whatever that prevents me from doing, it's not math!
I think there are three relevant questions:
1. What is the opportunity cost of isolating that one thing and drilling it?
2. How does that drill train your attention, muscles, and mind?
3. Will your new drill-refined habits help or hinder you in future situations?
/jesse-farmer.php?q=๐ retweeted
curious about the training data of OpenAI's new gpt-oss models? i was too.
so i generated 10M examples from gpt-oss-20b, ran some analysis, and the results were... pretty bizarre
time for a deep dive ๐งต
The LLM Catch-22:
1. You shouldn't anthropomorphize LLMs
2. Conversational UIs underpin their (consumer) popularity
as had been true for over a year, if youโre using subagents to play house and anthropomorphize roles, youโre hosed.
Claude subagents are context management + prompt shortcuts. That is all. That is all you need. If you wield them right they are incredibly powerful. There are engineers that are 100x more effective than you because they do this well.
Itโs not hard to learn but you have to tune out most of the hype.