Howie Liu's first startup, Etacts, was acquired by Salesforce about one year after it started.
“my first company lasted exactly a year, um, almost to the day, from start to acquisition, right, by Salesforce.”
Person
Howie Liu
1 episodes · 22 quoted claims
Airtable took roughly two and a half years to build its generally available product before public launch.
“Airtable, to your point, took probably 2 and a half years to even build the GA product that we could launch publicly.”
During a college internship at an ad tech startup, Liu built a custom asset manager for creative assets.
“I worked for this ad tech startup. And, like, the the, um, you know, I actually ended up realizing that the most valuable thing I could do during my summer there”
Howie Liu interned at CrowdFlower, Lukas Biewald's earlier company, which Liu describes as an early pioneer in AI training data.
“interning for your last company, CrowdFlower. It was kind of like early pioneer in in, um, uh, a lot of what would become this AI world and training data”
The Idaho Farm Bureau published an op-ed describing Airtable as the new CattleMax, a cattle-tracking software product.
“the Idaho Farm Bureau had a op-ed in it that referenced how Airtable was the new CattleMax.”
Liu says Airtable started around the same time as Figma, and both benefited from shifting user expectations and better web technology.
“we started around the same time as Figma, but I think what we both benefited from was both a shift in consumer”
Airtable built a web-native, desktop-grade product using modern standards such as HTML5 and WebGL, and its own proprietary database engine.
“using a lot of the benefits of, like, HTML5, WebGL, some of the, like, just modern standards and web”
Airtable did not launch with spreadsheet import, and added it soon afterward, along with templates, to keep growth from flattening.
“I don't think we launched with even the ability to import spreadsheets, but that was a fast follow, right?”
Liu says Airtable's biggest customers' value can roughly double every year.
“we can compound our our biggest customers value by basically 2x every single year.”
Liu says Airtable sees its best growth in retail, financial services, media and tech.
“we see the best growth in retail, financial services, media, uh, and tech”
Some Airtable customers pull data from Snowflake or Databricks into Airtable to build operational workflows.
“now we have a lot of customers who are pulling in data from, for instance, Snowflake or Databricks into Airtable”
Airtable is building a new data store to ingest data sets of several hundred million records.
“we're actually building a new data store into Airtable that will allow you t uh, allow you to ingest multi-hundred million record data sets into Airtable directly”
Airtable launched an AI co-builder that builds an app from a user's description, and lets users generate an app before signing up.
“we just launched, uh, our AI co-builder experience.”
Airtable reorganized its product group about nine months before the interview, moving from feature-oriented teams to teams organized around major pillars.
“we did this reorg in our product group, um, about, call it 9 months ago now”
Liu says context windows grew from a few thousand tokens two years earlier to 200,000, and up to 2 million in the biggest models like Gemini Pro.
“as the context windows have radically expanded from, like, you know, I think the best ones had a few thousand as of 2 years ago to now 200,000”
At the time of the interview, Meta's Llama 400 billion parameter model was coming out, and Liu said open source models appeared to be catching up.
“obviously, you know, very timely, uh, you know, to discuss this with the the Llama 400 billion parameter model coming out”
Liu says he sees upside in Adobe because Scott Belsky is working on its AI capabilities.
“I'm pricing a lot of, like, upside into Adobe, right?”
Tome started as an AI slideshow tool and pivoted toward a customer intelligence and presentation tool for salespeople.
“Tome, as an example, started as more of, like, this AI, uh, slideshow tool, right?”
Liu cites Webvan as arriving about a decade before Instacart, illustrating the importance of timing.
“Web van predated Instacart by at least what, like, a decade, and that timing was wrong”
Liu says that as a teenager during the dot-com era he signed up to beta test a smell-printing product.
“I signed up to be a beta tester for a smell printer company, right?”
Liu notes that Amazon emerged from the Web 1.0 era and became stronger after the bubble's rationalization.
“you had really good companies like Amazon that actually were born in that era, right?”
Liu cites a tweet about a customer whose main AI strategy was Microsoft Copilot but who could show no measurable return on it.
“somebody saying, like, I'm talking to a customer who, uh, you know, who is, you know, saying Microsoft Copilot is their main AI”