I’ve been seeing absurd efficiency gains with AI lately
But trying to understand why I was so surprised led me to a realization
When working with AI, it’s easy to credit the models for almost all of it. In reality, the wins came from a much deeper exercise in first-principles problem solving
It’s tempting to jump straight into the prompt box and feel like an efficiency god watching a model "think" for 15 minutes. Yet after a few stretches of idle waiting, I wondered: which parts of this process should be optimized or killed altogether?
Because of the immediate feedback loop with AI, we've stopped thinking about spotting basic overprocessing, waiting, or unnecessary handoffs, we blindly optimize for raw speed. But realistically, at least 60% of our 10x gains was just a classic continuous improvement exercise. AI simply handled the boring execution (triaging, tagging, parsing transcripts, etc)
There is plenty of hype around AI automation, but from what I can see, most of it is just putting a faster machine processing pure operational waste
2
239
this is a very neat explanation on how @link balance works!
The most satisfying way to understand how agentic payments work on @link👇
with @StripeCrypto, @tempo, @privy_io & @USDC
gaplab.me/agentic/link-balan…
1
179
Today I was lucky enough to see an Edward Hopper painting up close and couldn’t help but think that there is a fine balance between skill & being able to communicate an idea when it comes to making art
The thing is, some people mistake "intention" for "good art”
The thing is, there is a lot of bad art out there that people blindly accept as good just because it’s on a museum wall — they’re usually too afraid to speak their minds out loud for fear of retaliation from the herd.
If an art piece has to overexplain itself (Pollock) to prove its worth, that’s already an indicator that either the skill or the idea is missing
True art requires both at the same time
Please stop arguing in favor of bad art just because it has "intention."
Intention alone will never resonate with anyone, whereas an idea communicated skillfully (like Hopper does here) will be obvious to even the most casual observer
8
413
SpaceXSI Compile NYC was so much fun!
Apparently im in the top 8% of earliest @cursor_ai adopters, hence the badge 🫡
Friends who gift books are the best type of friends
Thank you @theesabina !
Gemini’s Argon does this annoying thing where it builds web artifacts for simple answers and it’s driving me crazy
Any ways to turn this off?
315
I cannot tell you how bullish I am on Stripe in the agentic era.
They are the ONLY connector inside Grok Bot that have the kind of granular permissions + human-in-the-loop prompts I *expect* for agents.
Every other connector - like ones from Google for eg Gmail or Docs - feel hopelessly amateur. Like WHY do I not have a way to have the Gmail connector ask for this kind of user permission before it lets an agent read any *sensitive* emails like password reset ones etc.
Stripe ygmi and I have no idea why others are not copying the very least what Stripe does
2
188
1B tokens processed today
Finally built an automated system to triage support tickets & route them to an army of independent agents built on specialized skills effectively reducing by 2x the time to investigate by human agents
This helped us reduce our backlog by 75% in just 4 days
Token count means nothing if it’s not tied to tangible process improvement (and happy customers)
2
9
541






































