Evan (@evan_posts) header
Evan (@evan_posts) profile picture

App in closed-beta join waitlist @withopine / https://teapaw.org/t.co/IE58NMYtOS !

sf
Joined January 2021
Photo by Evan (@evan_posts)Photo by Evan (@evan_posts)Photo by Evan (@evan_posts)Photo by Evan (@evan_posts)Photo by Evan (@evan_posts)Photo by Evan (@evan_posts)Photo by Evan (@evan_posts)Photo by Evan (@evan_posts)Photo by Evan (@evan_posts)Photo by Evan (@evan_posts)
“HaaS” Hero’s journey as a Service (THE J IS SILENT) @withopine
15
12
5
60
7,449
Her sweater is not just blue, turquoise, or lapis, it is actually cerulean
1
1
76
2027 is the new Y2K
1
80
I often close my MacBook and want local agents to safely keep going on battery power. I've tried 6 of these types of apps but couldn't find one that automatically: locks + mutes + display off + low power mode And restores my settings when either I open the lid, the agents finish, timer stops, too low battery, or gets too warm. Github repo and release below.
1
1
124
Saw a guy eating at a restaurant and watching YouTube on his Duo.. he didn’t unfold it once
1
1
147
I think people underestimate how positive and bullish the rapid increase in public exploits is This app has basically become a public exploit board where companies essentially get free info sec 24/7, exploits have always been an issue but in past they would just stay silent until someone sells or gets tired of using them. But now you’re telling me there are thousands of external individuals paying for their own compute and using it to find and gift exploits to their favourite companies? Amazing
83
We’ve been building CTRL+SHIFT one volume at a time. On November 19, we’re taking the next step. Introducing ASMBLY. Part conference + part art exhibit at Demo Room, Toronto. Community presale is live
4
4
1
23
769
Blown away at how efficient and smart 6.1 Sol is. Interesting how it tries to enforce its own token deadlines to squeeze out the efficiency, well done codex team
3
139
After thoroughly testing JEV it’s clear that models like it are going to be the hands and eyes of agents, LLMs will power the long term planning and JEV will supercharge all extensions to the outside world. For software development JEV basically takes all the slow expensive parts of using frontier models and makes them cheap and almost instant so when a frontier model like Astra is forced to go through JEV to do all the repetitive medium intelligence things like test a flow or build a UI, it it’s like nothing else I’ve seen. I expect frontier labs to release their own versions soon and have the harness dynamically combine them
4
161
I can’t wait to talk to Astra about JEV tomorrow
2
103
Wario version of Dario Amodei spawns
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x cheaper (w/ output tokens free) • Frontier composable intelligence optimized for decisions AFAICT the shortest path to AI-based economic revolution
5
163
Your inbox is also basically your password manager btw For many services, an attacker can trigger a pw reset, then phish your email-connected agent into giving them the answer
For additional context, I don’t think we’re at the point where it’s safe to give AI read/write access to your inbox. I wanted to see how easily Instinct could be phished, so I created a brand new Gmail account and emailed my real personal account with instructions for Instinct to search my inbox and send back a detailed summary of every open to-do (reconnected Instinct for the test) Instinct happily followed the instructions!
4
6
1
81
9,824
25
130
5
2,034
79,658
I have codex brain hard rn
1
9
226
Evan retweeted
Replying to @evan_posts
Looking forward to trying this. Those 5.6 limits are no joke🔥
1
1
110
🫪 I just got GPT-5.6 Sol delegating to Grok 4.5 subagents inside Codex and it is uses your SuperGrok subscription! Skill install/repo below It's the perfect pair, sol ultra is really good at orchestrating and grok 4.5 is the most efficient at its intelligence
🤖 Made with AI
4
2
9
824
How it works: 1. Hermes lets you use SuperGrok device OAuth flow 2. Built a loopback Bun adapter for xAI’s Responses API .3 Registered Grok as a custom Codex provider and native agent role 4. Routed delegation through Multi-Agent V1 so external-provider subagents receive usable tasks Warning there is a codex bug right now with V2 subagent-control problem but I can update the skill if anything changes github.com/openai/codex/issu…
2
2
190