Companies currently favor several use cases for generative AI. Among them, assisted software development occupies a prominent place. The main developers of GenAI models, such as Microsoft, Google, OpenAI and Mistral have positioned themselves in this market segment.
A French-American startup, Poolside, chose to dedicate itself entirely to this use case.
To finalize the development of its solutions and start the commercial phase, the Paris-based unicorn is raising capital.
Over 100 million developers to convert
After an initial funding round of $100 million in 2023, Poolside has just formalized a Series B totaling $500 million. For the generative AI giants, investors are not hesitating to pull out the checkbook despite the context not being very favorable for startups.
OpenAI signed this week a record deal of 6.6 billion dollars for more than 150 billion valuation. For its new fundraiser, Poolside appealed mainly to the Anglo-Saxons.
The Series B was actually led by Bain Capital Ventures. DST Global, StepStone Group, Citi Ventures and HSBC Ventures are also participating, according to Bloomberg. The young company now has a valuation of $3 billion.
Poolside has at least one strong argument to convince investors: the size of the population covered by its coding assistant. In an interview, its co-founder Eiso Kant reports that the number of developers worldwide exceeds 100 million.
AI will overtake humans in terms of code for Poolside
Therefore, the market for IAGen applied to code looks promising in theory. However, in order to generate profits, the startup is not targeting all developers. Manager is primarily aimed at customers who employ 5,000 engineers or more.
Even if Poolside does not yet have a commercial product, its founders show their optimism in a post published on the publisher’s website. For Jason Warner and Esso Kant, the superiority of machines over humans in coding is only a matter of time
“It is precisely because of data and computational limitations that we believe software development will be the first high-impact opportunity where AI will match and exceed human capabilities.” write-ils.
The way to achieve this, they say, is based on code execution feedback reinforcement learning, code execution feedback reinforcement learning, or RLCEF.
$500 million for RLCEF scaling and model training
“Our models explore solutions to millions of tasks across 130,000 real-world codebases (and growing), where they get runtime feedback on every solution they try to implement,” the founders explain.
Poolside will need solid assets in the face of increasing competition in the code generation world. That’s why the startup’s ambition is to offer models that reach “human capabilities in software development.”
To do this, he needs to push his reinforcement learning method. Funds provided by investors are a direct response.
“Thanks to this new fundraising, we were able to grow our training cluster to 10,000 GPUs and begin work on scaling RLCEF and training our models,” the leaders say.