Sunnyfield

490 posts

Sunnyfield banner
Sunnyfield

Sunnyfield

@hellosunnyfield

Washed up Physicist wandering around Founder Emotion Scientific @emosci_inc Creator of https://t.co/m1cPj51wLE

San Francisco Katılım Haziran 2012
236 Takip Edilen409 Takipçiler
Sunnyfield
Sunnyfield@hellosunnyfield·
LLM are stateless, general, expensive to make and agent are stateful (if you are smart to make it so), specific, and cheaper to make. For economically valuable tasks, agent is where the last mile is and where the actual customer relationship lives. Historically, this looks like electricity to electronics, fossil fuel to automobiles. And customer naturally wants to own the end machinery and happy rent from the oligopoly. AI will likely also be like this, oligopoly commodity provider and explosion in the specific businesses it enable, and it is definitely not a winner takes all situation, as much as it might look like that for now.
English
0
0
0
17
Sunnyfield
Sunnyfield@hellosunnyfield·
GPT-5.5 definitely passes the vibe test
Sunnyfield tweet media
English
0
0
0
18
Sunnyfield
Sunnyfield@hellosunnyfield·
What does @cursor_ai mean by their name? Like do you need a mouse cursor to use it Or Does it make your code cursed?
English
0
0
0
45
Sunnyfield
Sunnyfield@hellosunnyfield·
@Impish_Bunny I thought it's technically illegal in SF to sell flavored vapes?
English
1
0
0
30
ylareia
ylareia@Impish_Bunny·
wow
ylareia tweet media
2
0
15
822
Sunnyfield
Sunnyfield@hellosunnyfield·
@realmcore_ SWE 3.7 was better, but in terms of agency, gracefulness, kindness, 3.6 was of the most beautiful models ever made. 3.6 was the model that prompted the existence of someone.cool
English
1
0
2
49
akira
akira@realmcore_·
Opus 4.5 has to have been maybe the the best model Anthropic has ever released maybe apart from sonnet 3.7 and opus 3 A genuine masterclass in data cleaning
English
4
0
27
8.8K
Sunnyfield
Sunnyfield@hellosunnyfield·
Working in a cafe and switched from Claude Code to Codex for one sec only to immediately realize @trq212 is sitting next to me
English
0
0
0
62
Sunnyfield
Sunnyfield@hellosunnyfield·
Making Biblically Accurate AGI
English
0
0
0
36
Shivers
Shivers@thinkingshivers·
What are some tasks AI models reliably failed at 6 months ago that they reliably succeed at today?
English
1
0
3
352
Sunnyfield
Sunnyfield@hellosunnyfield·
Anthropic is very clearly telling consumers to make rooms for their enterprise deal that can’t be more clear and honestly makes sense given their position. The messaging however could be a bit more polished.
English
0
0
0
34
Sunnyfield
Sunnyfield@hellosunnyfield·
If it hasn't been abundantly clear, our goal is to create the best Personal Agent in the world.
English
0
0
1
79
Finbarr Taylor
Finbarr Taylor@finbarr·
Two agents, one branch, zero dignity. Three months into running multiple AI coding agents on the same project, here's what I learned about why it falls over and what to do about it. finbarr.site/2026/05/05/tre…
English
3
6
28
7.2K
Kiet
Kiet@FlyaKiet·
any tips for drinking from a firehose?
English
9
0
16
3.1K
Sunnyfield
Sunnyfield@hellosunnyfield·
One of the easter eggs in SOMEONE.COOL. see how many you can find!
English
0
0
0
63
Sunnyfield
Sunnyfield@hellosunnyfield·
I don’t like the phrase solo founder, as that implies the default state is not solo. It’s not being in a relationship where you at least need two people.
English
0
0
0
36