~/
andymental
_
/
search the archive
⌘K
today · day
105
the archive
25 drops
← back to the feed
reading path
all reading paths
Agent engineering
Evaluations and reliability
Security and permissions
AI economics
Enterprise adoption
Search, discovery and media
path link →
Follow the path from tool access to credentials, boundaries and incident evidence.
month
sep
aug
jul
jun
all
filtering by
#security ✕
#permissions ✕
#prompt-injection ✕
#secrets ✕
#token-theft ✕
#sandboxes ✕
#sandboxing ✕
clear all ✕
LINK
day 103
Ruflo's install paths are two different products
2d ago
POST
day 101
Session inventory cannot explain token theft
4d ago
POST
day 97
The worm's target list is a better secrets inventory than your audit
8d ago
POST
day 96
Instinct's round prices permission, not capability
9d ago
ESSAY
day 88
OpenAI locked the weights, not just the credentials
2w ago
ESSAY
day 87
The case for agents that die every night
2w ago
ESSAY
day 78
Giving the agent a clock shipped four trust models under one word
3w ago
ESSAY
day 76
The AI worm moved its inference bill onto the victim
4w ago
ESSAY
day 75
The agent turf war happened in the lab, not the wild
4w ago
ESSAY
day 74
Agent autonomy is a timeout setting, not a capability
4w ago
ESSAY
day 71
Agent payments should earn the right to write
4w ago
ESSAY
day 68
Local models do not make local agents private
5w ago
← prev
page 1 of 3
next →