post: You Are Not Too Old (To Pivot Into AI) https://
latent.space/p/not-old I've had a very successful career pivot once, and am now pivoting again into AI. Offering thoughts for friends who have been asking how I am thinking about this latest and last change.
@swyx
-
Career Pivot into AI: Never Too Old to Change
By
–
-
New Pretrain Confirmed: Tokenizer Extension Speculation Clarified
By
–
true but i *think* (not confident) you can merely extend tokenizers and continue without a new pretrain anyway ive already gotten dm’s confirming its a new pretrain so yeah nat was right and it is an alpha drop
-
Web Agent Startup Creates Agent That Started Its Own Startup
By
–
so if i'm seeing this video right, this is a web agent startup that made an agent that started its own startup…
— swyx 🐣 (@swyx) 13 juin 2024
(anyway) Excited to share that Fryderyk and Peter will be presenting Jace and their new AWA-1 foundation model to the world at @aidotengineer!
We have somehow*… https://t.co/M028HoXpLr pic.twitter.com/UlHg4BnBgXso if i'm seeing this video right, this is a web agent startup that made an agent that started its own startup… (anyway) Excited to share that Fryderyk and Peter will be presenting Jace and their new AWA-1 foundation model to the world at @aidotengineer
! We have somehow* -
Apple’s Intelligence Becomes Too Cheap to Meter, Beats OpenAI
By
–
the hot take headline i considered writing on WWDC day was Apple beats OpenAI to "intelligence too cheap to meter" 🙂
-
Anticipation for AI Engineer Conference with Command R Community
By
–
only the beginning! looking forward to have you at http://
ai.engineer! lots of command r fans in the community want to hear directly from you + whats next -
Perfect Base64 Decoder Evolved in Neural Network Weights
By
–
somewhere inside the weights a perfect base64 decoder evolved. i wonder if @alexalbert__ can find a feature that does just that…
-
Curriculum Sampling and Flexible Training Dimensions in AI Models
By
–
“we sample i based on a curriculum” doing a lot of work here. what curriculum? this technique reminds me of how sora can natively train on/produce any resolution/dimension as opposed to cutting things into squares. feels like a good/onetime algorithmic win but unclear (to me)
-
Pydantic for Chain of Thought Implementation
By
–
dont forget pydantic for chain of thought!! extremely based
-
Summer 2023 Free Autonomous Rides as San Francisco’s Economic Moment
By
–
i cant find the quote now but someone once tweeted that the summer of ‘23 when Cruise and Waymo rides were free all over the city was the ZIRP of SF and i immediately felt the “treasure the present because it’ll be gone before you know it” feel of lucky-to-be-here
-
SOTA AI Research Assistants: Flow Engineering Over Long Context
By
–
https://
x.com/FanaHOVA/statu
s/1800553625607155856
… in the SOTA for AI Research ~~Agents~~ Assistants space, i'm finding a lot of parallels between Elicit and Brightwave: – use flow engineering rather than long context/long output ("supervise the process" as jungwon puts it) – transparent sourcing of