Many startups and larger tech companies have taken a crack at building artificial intelligence to code software. Now, another new player is coming out of the shadows to throw its hat into the ring, with a mission to fix the many problems that will arise when humans and all those AIs are all writing code together. Tessl is building what it describes as an “AI native” platform that developers and their teams can use to create and maintain software, and it’s today opening up a waitlist for those interested in trying it out.
“Is building” is being used very specifically here: Tessl’s product has yet to launch (the plan is to have it out by early next year). But the London-based startup is sharing a bit more about what it’s doing with some financial fanfare. Tessl has quietly raised $125 million across a seed round and a Series A, both being announced for the first time today. The latest round is led by Index Ventures, with Accel, GV, and boldstart participating. GV (aka Google Ventures) and boldstart co-led the seed.
TechCrunch has confirmed with multiple sources that Tessl’s post-money valuation is north of $500 million.
As you might have surmised, one reason why a company without customers nor a shipped product is getting this kind of attention from top-shelf VCs is because of who is building it.
The CEO and founder is Guy Podjarny, a developer whisperer of sorts. His last startup was Snyk, a cybersecurity firm that was last valued (in 2022) at $7.4 billion. Before that, he was the CTO of Akamai, a role he took after Akamai had acquired his first startup, Blaze, which focused on speeding up website loading times.
“Podjarny is incredibly visionary and thoughtful about his business,” said Carlos Gonzalez-Cadenas, the partner at Index who led on the investment, in an interview with TechCrunch. “He’s very, very good [at understanding] developer communities and building developer oriented businesses.”
Podjarny said in an interview that the concept for Tessl came out of his experience at Snyk.
The latter firm’s focus is on detecting (and fixing) security vulnerabilities in code, and Podjarny saw a similar issue getting more urgent with code and software interoperability overall — in particular because of the rapid expansion of code written automatically by AIs.
“What is AI doing to software development?” he said he asked himself. The answer was: speeding it up, but also creating much more of it automatically. And the process of maintaining and shipping updates to that code would compound the complexity and chances of systems breaking. This ends up having a lot of bad implications (security, uptime, cost, efficiency) for organizations. “The more that picture formed in my mind, the more I knew I would build this.”
(The name Tessl is a reference to “tessellation”, Podjarny said: the aim is to make sure that software and the code behind it fit neatly together, rather than exist in a messy, overlapping jumble.)
Podjarny was cagey about giving out too many specifics about what kinds of applications or code he envisions being built, or maintained, on Tessl. But it sounds like while the ambition sounds big, it will actually start small.
“We’re not sharing the full strategy yet on on what that is,” he said of target applications or use cases. “I would say that we’re not starting with games. We’re starting with relatively simple software that that allows us to build an end to end system that is more manageable for LLMs to generate, and more manageable for humans to specify. And we will evolve from there.”
The basic idea behind what the startup is doing goes something like this: developers and their teams (which can include product managers and others that typically do not write code themselves) can provide specifications to Tessl. These can be in the form of natural language, or code itself. Those specifications can then be used to Tessl to write new code to meet that specification. You and your team can also test that code in a sandbox, where issues can be flagged and addressed, and continue to modify specs as needed. After that, Tessl can be automated to maintain that code to that specification: meaning, if something else runs the risk of breaking because of that new code, Tessl will run remediation to identify and fix that.
It sounds like Tessl is not being conceived as a walled garden. Podjarny said that he’s talking to others that have built or are building AI coding assistants, with the idea that the work these other platforms produce will also be maintainable using Tessl. It will aim initially to support Java, Javascript and Python, with further languages getting added over time, Podjarny added.
One reason why investors love the idea and are backing it is because of that extensibility. Maintaining code is something that he said had “a lot of signal” as being important right now, said Gonzalez-Cadenas. “But he’s building a system of record here,” he added. “Once you do that, there is a variety of opportunities.”
