Pinned Tweet
Join us today for an AMA with one of the most respected engineers in the industry who recently gained the legendary "Distinguished Engineer" title at Google! @kelseyhightower on Twitter spaces 👉🏿 teapaw.org/i/spaces/1kvJpmLywvXxE
Esco Obong retweeted
Atlassian: "hey can we talk?"
Me: "Remember when you acquired DX for $1B? And the CEO canceled my advisory shares and the advisory shares of a bunch of other advisors. And screwed over a bunch of employees on the sale.
Until it's sorted for everyone, no."
Esco Obong retweeted
Why did GitHub Copilot have no response to Claude Code? @satyanadella says:
"Code went from input to output effectively. That was the transformation. In six months or so, IDEs faded away.
When model jumps happen non-linearly, the form factor has to change quickly. AI-first product building is unforgiving because you have to build everything around a given model, then be ready for the next model to show emergent capabilities and rewire yourself for that.
Don’t be in love with what you built."
HOT TAKE: Surge pricing is actually good and things would be worse without it.
Surge pricing is how Uber attracts more drivers to an area with high demand.
Uber sends notifications to drivers about the price increase, even ones who aren’t working at the time. This gives them a reason to flock to the area where people need them most.
The goal is equilibrium not price gouging. To serve as many customers as possible you need more drivers.
Uber can’t force drivers to go anywhere. Keeping the prices unchanged, keeps the supply of drivers unchanged. And even worse, drivers would avoid hotspots because of the chaos and bad traffic.
Paying more increases the supply and actually brings surge pricing back down once enough drivers flock to the area.
That’s different from airlines hiking fares or vendors raising umbrella prices during a storm. None of those solve any problems, they just charge more. The plane has no extra seats, and the vendor has no extra umbrellas.
What’s the alternative solution to increasing the amount of drivers in an area 10x more than normal?
I think for a humanitarian crisis where safety is a concern, a good solution would be to have a fund that pays for the increased cost on behalf of customers.
Because you still can’t avoid the need to pay more drivers, but you can plan ahead of time to subsidize certain events, like some sort of insurance policy.
201
LinkedIn is literally a drug. I remember when it went from an online directory to a legit social media platform. I’ve been writing code long enough to know none of this happened by accident.
LinkedIn ran the social media playbook laid out in Hooked by Nir Eyal.
When they added the feed, the goal was to build a habit loop.
A notification pulls you back into the app. They measure how you engage with posts in your feed, learn what works and use AI to optimize for posts most likely to keep you coming back.
The goal is to create dopamine-driven anticipation, similar to drugs. Your brain keeps looking for the next reward.
Eventually opening LinkedIn stops being a decision.
148
A look into the future of security from saedsaed Senior Security Engineer at Google, and Esco Obong, Senior Software Engineer at Airbnb.
Everyone is saying SaaS is dead.
But the biggest value in buying services is maintenance. People may trade off reliability for convenience, but security is a deal breaker.
It’s one thing to build an application with AI in a weekend. It’s another to monitor vulnerabilities, patch dependencies, respond to incidents, and keep everything secure for years.
SaaS companies will remain if they offer a team dedicated to providing a level of security that individuals can’t achieve or keep up with.
Exploits like PHP2Shell and HTTP2Shell are perfect examples of why.
These vulnerabilities exist deep in the technologies that applications depend on. You could write perfectly secure application code and still be vulnerable because of something running underneath it.
And with AI accelerating vulnerability discovery, the time between vulnerabilities being discovered and attackers exploiting them keeps shrinking.
Most individuals and small teams aren’t equipped to patch security flaws rapidly.
The future of SaaS won’t be about selling software that people can’t build themselves.
It will be about selling software they don’t want to be responsible for securing and maintaining.
What do you think?
1
4
266
Esco Obong retweeted
"god when is meta ever going to launch muse voice they announced it forever ago"
muse:
🤖 Made with AI
Esco Obong retweeted
Elon Musk took a $40,000 username from a guy who had owned it for 16 years.
He paid him absolutely nothing.
Not even a dollar.
Instead, he was offered a T-shirt and a tour of Twitter’s headquarters.
This is the story of Gene X. Hwang the first @X username
2007
- photographer from San Francisco
- co-founder of Orange Photography
- registers @x on Twitter
- keeps the username for 16 years
over the years
account grows to ~53,000 followers
receives offers of up to $40,000 for the handle
But he refuses to sell
2014
- hackers steal his x account
- Twitter helps him recover it
July 23, 2023
- Elon Musk announces Twitter will become X
- Gene is in Canada playing in a pinball tournament
He has no idea what’s coming
July 25, 2023
- Twitter officially becomes X
- the company wants x
- Gene receives an email from Twitter support
“Your user handle is affiliated with X Corp.”
his username is taken and gets a new one @x12345678998765
what X offers him
- company merchandise
- exclusive tour of X headquarters
- meeting with the management team
$0.
Gene even asks if he can have the old blue Twitter bird sign being removed from headquarters.
they say no.
October 2025
X launches its Handle Marketplace
rare usernames can cost $2,500 to over $1 MILLION
users must maintain a paid subscription
lose the subscription, potentially lose the handle
the same company that took x without paying its owner now charges people for access to rare usernames.
16 years building an identity.
$40,000 in offers rejected.
and all he got was an offer of merchandise and an office tour.
BROTHER.
Esco Obong retweeted
Credit to Dario who said pretty much exactly this would happen, and on this timeline, thanks to seeing what Mythos could do.
In May, Dario said there's a 6-12 month window before Chinese models catch up was also correct. That was 5 months ago.
I don't push code anymore. I deploy prompts.
Every production server has an agent process always running. It reads the prompt, writes the code directly to disk and reloads the app server.
No pull requests.
No code reviews.
No CI.
Just pure execution.
Time between code generation and serving users is instant.
Each production instance generates slightly different code — but they all have the same goal.
Continuous Deployment is too slow.
Welcome to the era of Vibe Execution.
3
202



















