ShrekOverflow's profile picture. Product Architect @Auth0 for AI Agents, decade+ in #iam #security #privacy #oauth2 #oidc. Posts about security, agents, large systems and structures

ShrekOverflow

@ShrekOverflow

Product Architect @Auth0 for AI Agents, decade+ in #iam #security #privacy #oauth2 #oidc. Posts about security, agents, large systems and structures

置頂

Predictions for the next (n) years. Companies that fail to innovate will see AI as a cost cutting measure. Companies that innovate will see it as an accelerator. I think @luke_lafr said that about 1 year ago. The has been the most accurate prediction so far 🔮 clear…


So wait a minute, the M4 Max studio is cheaper, 2x power efficient and performs better than the DGX Spark? MLX[cuda] is some 4d chess esp if the ~5x prompt processing carries over to M5 Max.


Hey erm @1Password have we always been able to search for passwords in the search? Or is this a new bug?


After trying ghostty am wondering why didn’t I do that earlier. Taste, Polish and attention to detail are such a hard differentiator


Hey folks at @vercel can I please have v0 app?


Turned a Mac mini (via Linux) into a KVM Switch (TB only) anyone want to try?


Getting to the $20 for free consultation levels now @claudeai :'(

ShrekOverflow's tweet image. Getting to the $20 for free consultation levels now @claudeai  :'(

At 60-80% of context window the model performance becomes really "lazy". Feels like a person at the end of the workday, whose building coffee for models?


The most idiomatic way to use @cursor_ai 's plan mode seems to be. 1. Vibe code, back and forth in Agent Mode, get it to work, 2. Then have plan mode take over fresh, and use it to plan and then build the next version, from previous scratch stuff. 3. Now plan mode -- again,…

ShrekOverflow's tweet image. The most idiomatic way to use @cursor_ai 's plan mode seems to be. 

1. Vibe code, back and forth in Agent Mode, get it to work, 

2. Then have plan mode take over fresh, and use it to plan and then build the next version, from previous scratch stuff. 

3. Now plan mode -- again,…

If Nvidia is using Cursor, there is no reason why any other enterprise can't use Cursor, for "compliance and privacy reasons"


Been my way to build basically anything since even before AI. Though creating a culture and sustaining it is the real “cracking”

This is how Anthropic decides what to build next—and it's brilliant. Instead of endless spec documents and roadmap debates, the Claude Code team has cracked the code on feature prioritization: prototype first, decide later. Here's their process (shared by Catherine Wu, Product…



Taking time off and falling sick + going to through a large back log sucks 😅.


What does login actually mean? It roughly boils down to proving you are who you claim you are.


Random appreciation post, @AmericanExpress really cooked with the mirror card. Damn it I need to protect this thing


Economics always wins folks, always. so happy @libuild bet on pickaxes


See, MCP _is_ an app layer, not an API layer, apps built for chatGPT without care for context engineering are going to unleash a new kind of terrible experience 😂


Calling it already Agent workflow builder will eventually just JIT these workflows. The workflow topology is extremely high value in being able to build these ephemerally like humans do


OpenAIs AgentKit, Apps SDK are incredible building blocks that turn ChatGPT into an operating system 😁❤️


My current feed is filled with Hikaru yay or Hikaru nay. Remind me what game does he play? @GMHikaru is literally playing psyops chess and winning both the game and the feed


United States 趨勢

Loading...

Something went wrong.


Something went wrong.