Intelligence and Interfaces Eng Lead @CashApp
Joined June 2009
- Tweets 7,684
- Following 544
- Followers 1,571
- Likes 10,434
196 Photos and videos
We are uncovering better ways of developing software by doing it and helping others do it.
Through this work we have come to value:
- Individuals and interactions over processes and tools
- Working software over comprehensive documentation
- Customer collaboration over contract negotiation
- Responding to change over following a plan
That is, while there is value in the items on the right, we value the items on the left more.
1
11
2,440
Meet Cash App Mobile*: unlimited 5G on AT&T's network for $40 a month, taxes and fees included.
No contracts, no credit checks, and no store visits. Manage your phone plan right next to the rest of your money. Rolling out now 💚
cash.app/download
164
188
1,339
2,251,188
gertig retweeted
Jun 11
introducing ... < CASH APP MOBILE >
unlimited 5G for $40/mo for ... everyone
no long term contracts or credit checks
could NOT be more excited about this one
more to build --->
cash.app/press/cash-app-mobi…
61
57
620
569,312
Had a conversation with someone and they asked me how an agent knows when to stop working on something.
This felt very much like a curse of knowledge moment.
I explained that agents exist because models gained the ability to do tool calling, and that we use the lack of any tool calls as an indication to the harness that the model is done working.
What would you have said?
1
6
752
huge if true
May 5
Introducing SubQ - a major breakthrough in LLM intelligence.
It is the first model built on a fully sub-quadratic sparse-attention architecture (SSA),
And the first frontier model with a 12 million token context window which is:
- 52x faster than FlashAttention at 1MM tokens
- Less than 5% the cost of Opus
Transformer-based LLMs waste compute by processing every possible relationship between words (standard attention).
Only a small fraction actually matter.
@subquadratic finds and focuses only on the ones that do.
That's nearly 1,000x less compute and a new way for LLMs to scale.
1
272
This guy gets it
Apr 29
my morning psychosis - ralphing a better github on top of @freestyle_dev git server
new loop - I have heard from a few folks that codex 5.5 is the first model that can reliably simplify code without out messing it up, so we're doing 3 passes per loop:
codex 5.5. low - ralph
opus 4.7 high - ralph
codex 5.5 low - simplify this code, make @martinfowler proud
Repeat
stay tuned and follow along (link in comments)
4
757