Staff backend engineer @ Hinge ex-etsy, ex-@google ҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓̕҈̒̓

New York
The big divide is that a lot of AI companies also have super demanding hours, by virtue of being startups and fighting for position in the market. Feels like the job market is segmented into (Money + AI company) vs (Time + traditional company with a hybrid execution model) now
Today's management tip is that if you aren't an "AI company," all of your employees are currently considering whether they should leave you for an "AI company." "AI company" is defined as "posts on X / podcasts all the time about AI"; it is not a revenue or traction metric.
61
Jacob Voytko retweeted
BREAKING: Anthropic has filed for an IPO, per Reuters
Community note
Anthropic has not filed a public IPO registration. Reuters obtained and exclusively reported details from a non-public prospectus; no Form S-1 for Anthropic PBC appears on SEC EDGAR. reuters.com/business/finan… anthropic.com/news/confident… sec.gov/cgi-bin/browse…
218
423
6,177
800,323
Interestingly, I just got my first winback email outreach from Anthropic almost immediately after Sonnet launched
1
64
And back in the '00s, you did a phone screen by reading someone a problem out loud over the phone. Then they'd pencil-and-paper it, and then read the solution back one character at a time.
The new generation will probably never know I am wondering when most places (not Google) will drop Leetcode
51
OpenAI dev days predictions - $500 plan. 3x usage the $200 plan and Tibo has to respond to one of your tweets per month - Felony mode: run a swarm w/o guardrails - The world's buggiest desktop app. In 2 weeks you will learn that your company's analyst runs their whole life on it
1
34
This is because building your own Tag is easy, and might be better suited to your environment if you can write it in your company's language using your company's libraries.
it has been a few months now and nobody except ant employees seem to talk about claude tag
1
58
Trying my first "and rewrite it in Rust" port of my now-9000 line Python vibecoded workspace switcher I use for side project stuff I love it but god it's getting slow
1
34
"Why don't you just read what it does and fix it" that's why I'm making Astra rewrite it in Rust! So that I can't even if I wanted to. Everyone knows that nobody knows how to read or write Rust
16
Jacob Voytko retweeted
That DHH saying out loud that the age of writing code by hand is over for the industry, and causing such a huge uproar is a bit interesting, given this was clear enough since ~January for most of us using the models + paying attention. Removed paywall: newsletter.pragmaticengineer…
148
225
2,580
175,300
kinda surprised to see the sentiment on ChatGPT Sol 6 being so negative on Reddit... I had it tackle some feature work yesterday for my side project and do a performance investigation and it aced both. Is it bad at specific types of tasks or something?
67
gpt-6-sol and gpt-6-luna just appeared in my personal Codex account That didn't take long!
76
The next Jev will be a time series classifier. People will use it as an edge detector to activate ops agents to investigate abnormalities.
1
27
Also wouldn't be surprised if the current Jev could do this in some way, but I've seen enough people post weird numerical Jev results to be skeptical
1
16
"LLMs can do this" anything can poll, I'm talking about adding epoll baybee
4
Way off topic but I recommend doing what that L7 manager did; if you can't do a project in a week or two it's very likely your intern is being set up to fail with a project that's way too large for them
this is like when my L7 manager at google did my internship project the week before i joined but let me spend the entire summer solving it anyway
67
Happy /fast day to all who celebrate
Ladies and gentlemen... start... your... ENGINES. We are almost Tuesday and I promised a reset for Tuesday. Among some other things. See you soon.
2
96
Bear case for OpenAI/Anthropic: Anytime I ask myself "Could Jev help us avoid or reduce $huge_amount in LLM spend?" the answer is usually "oh, most definitely"
1
33
Jacob Voytko retweeted
GraphQL exists because getting the backend team to add three fields to an endpoint was harder than inventing a query language.
44
160
3,584
125,109
After using a lot of Opus, Fable, Sol, and Astra, it really feels like Sol is the only model you could just default to and never need to change it when on the $200 plan
2
50
Scrum is a reasonable default for organizing a new team (it's way better than "no organization at all"), but the team's not functioning at a high level if it never grows out of / customizes Scrum to meet its own needs
Scrum made sense at companies/teams that shipped a new version of their product every 2-3 months (or less frequent) As soon as a team got to shipping daily or had CD: they only held teams back Most startups + Big Tech moved on a decade+ ago, legacy companies remained only
1
2
106
A strict implementation of Scrum makes dozens of decisions about how work should be done. The odds that every single facet exactly meets your team's needs is approximately zero
1
1
22
The real difference is whether your team takes retros seriously. If you've been working for 3 months and you can't name a significant process change you trialed from a retro (success or failure), your retros have been a waste of time
13