"that walled garden turns into a leaking pipe" - so ppl are using AI to write about AI ruining businesses, how interesting
Airbnb is dead. Well it’s not dead but it’s going to be dying soon.
The Bear case for Airbnb.
I’ve been bullish on Airbnb as a company for many years. But my outlook on the business changed this week.
Let me tell you why.
In 2012 I purchased my first Airbnb property. Over the years that one property has turned into a portfolio of high-end Airbnb’s that generates almost $1M a month in rental income.
Using ai agents for the first time this week has completely changed my outlook on the company.
Airbnb is facing an existential threat that most of the market is completely missing.
With Airbnb pushing its mandatory host-side commission to 15.5%, the middleman tax has never been higher.
Enter AI agents. (Hark, Dot, Muse, Grok Bot, etc)
Right now, booking a vacation rental requires navigating a siloed app and paying heavy platform tolls.
But your AI agent won't care about Airbnb's brand wall. When you know the property you want to book and you ask it “what’s the cheapest way to book the property” it will query property management system APIs directly, find the exact same home, and execute a direct booking that saves everyone 15%.
Airbnb's entire model relies on holding supply captive behind a closed interface.
When AI agents can autonomously search, verify reviews, negotiate terms, and book direct with property operators, that walled garden turns into a leaking pipe.
The middleman tax's days are numbered.
I’m predicting that Airbnb’s growth rate will slow down significantly in the next 24 to 36 months. And within 5 years they will be seeing negative growth.
The same can be said for most other online marketplaces that don’t have unique inventory that can only be booked directly on their platform and nowhere else.
youtu.be/dHKhYL3is5o people working in Amsterdam Meta office don't like Meta products
Is "Lyle′s Variety Store" song from the Stellar Blade OST somehow inspired by Lyle Mays? I never played the game but the song sounds a bit familiar for the long time Pat and Lyle fan.
Eugene Lazin - @lazin.bsky.social retweeted
it still blows my f* mind every time @redpandadata hits production and a CIO texts you "You improved my home grown ETL 1000x" ... as we've added sql, streaming, connectors the impact basically compounds on prev layer. this kind of stuff gives me unlimited energy.
#redpanda
Interesting thread, but I think that one have to squint too much to see resemblance. Before S3 and clouds the most common type of disaggregated storage in a datacenter was SAN. But it's so different to S3 it's a meaningless comparison.
Eugene Lazin - @lazin.bsky.social retweeted
Hot Chips '26: NVIDIA says the New York Stock Exchange processes over a trillion records a day and sees a 6x speedup testing its new Vera CPU
"I want to give an example of some of our scaling. There's a collaboration between the New York Stock Exchange and Redpanda, and they process over a trillion records and 500 billion-plus messages a day across their seven exchanges."
"And Redpanda is a real-time streaming engine built for high throughput. And they're seeing a 6x improvement in some of their testing on Vera versus a traditional CPU."
"And this is all possible because our fabric lets all of our cores communicate and process data without any bandwidth constraints."
Head of Nemotron Bryan Catanzaro says NVIDIA is sitting on internal "teacher" models that beat everything it has publicly shipped and he's trying to release these as well
"This technology of using teacher models and multi-domain on-policy distillation has been really key for us in these ways."
"Now, what does that mean? Well, it definitely means that inside NVIDIA we have a collection of teacher models that are individually better at different domains than any of the models that we've released."
"And actually, I'm trying to release those teacher models as part of the Nemotron project as well -- because as you know, Nemotron is not just a model."
"Nemotron is an open technology effort. It includes data sets, it includes techniques, all of our ideas about how to build a model, which we publish."
"Of course it includes the models themselves, but I think it should also include collections of teacher models."
"I'd imagine this could be a really great way of federating open technologies for AI across the ecosystem."
_______
Wrote more on a prior Bryan interview here: firesidealpha.substack.com/p…
I can't live in my house for a few days already because my sewers pump broke. This is how I learned that a) sewers pumps exist b) my 120yo house has one.
The bloody thing is not even in my house. This is a row house and I'm occupying two top floors and my neighbor is occupying the rest. The pump is below his ground floor and he didn't know about it either.
Eugene Lazin - @lazin.bsky.social retweeted
We've been working alongside the @GlobalFoundries team, and today we get to share what we've been building together. 🤝
Together, we're innovating to build a single real-time platform that unifies equipment, process, and scientific data across GlobalFoundries' global operations. That same platform now powers a growing portfolio of AI agents across IT, procurement, and more, all governed through Redpanda's Agentic Data Plane. 💪
Read the full story →redpanda.pulse.ly/exdeq0gomm
Eugene Lazin - @lazin.bsky.social retweeted
I’m looking for my next role.
I’m a software engineer and product designer with deep experience in databases, sync engines, frontend systems, performance, and UI/UX.
Open to full-time or contract work, remotely or in Berlin.
Details in reply ↓
RT for reach 🙏




















