AI coding funding frenzy…down rounds (thank goodness) declining…Miami VC loses its luster

Main Item

It’s been a tough week for generative AI coding startups.

As the industry returns from Burning Man and Labor Day, activity is heating up fast.

Codeium and Magic, two coding assistants that have developed their own underlying models, announced this week that they have joined General Catalyst, Sequoia, Nat Friedman and Daniel GrossIn the spring, coding assistance company Augment raised a massive $277 million in Series B funding from Index and Lightspeed Venture Partners.

Introducing Cursor, an AI code editor that can adapt to your coding style over time $60 million Cursor received Series A funding from a16z, Thrive Capital, and OpenAI at the end of August. Claude 3.5 Sonnet ModelAnthropic’s models consistently rank among the best in various code-writing benchmarks, and three VCs who weren’t involved in the Cursor round told me that the combination’s superior features stand out.

The founders are now looking to build more sophisticated coding agents that can work across platforms to solve complex problems. Replit, whose “Core” subscription costs $10 a month, launched on Thursday. Release The company says its newest agent can write code, build a developer database and deploy software based on plain English prompts.

Another startup, Honeycomb, operates multiple agents fine-tuned for different developer tasks, like bug fixing and code testing, and can integrate workflows across systems like GitHub and Jira. The company is in the final stages of Y Combinator’s Summer ’24 session, which airs this week. top SWE bench LeaderboardThis measures how accurately the model can solve real software problems on GitHub.

The whirlwind of activity over the past few weeks proves that the field is one of the first to demonstrate rapid enterprise adoption and big revenue in generative AI. Unlike many other industries, where adoption has been slow due to concerns about accuracy, cost, and regulatory and legal compliance, software developers are using AI coding tools. a lot already.

“This is one of the few areas in the enterprise where we’re seeing real value,” one GP told me backstage.

Not everyone should invest in venture capital, but everyone should option Invest in venture capital.

Or so they say. Fundrise Innovation Fund.

The pioneering fund has already raised more than $125 million in an effort to level the playing field in one of the asset classes still largely untouched by the retail revolution.

Its portfolio includes some of the biggest names in the technology industry. Fundrise Innovation Fund Easier than ever Anyone “Get in early”

View Portfolio

Please consider the investment objectives, risks, charges and expenses carefully before investing. This and other information is contained in the Innovations Fund prospectus at Fundrise.com/Innovation. This is a paid advertisement.

According to a new study by GitHub published on Wednesday: 97% of developers Survey participants have already used AI coding tools at work, whether or not their employers explicitly allowed it. Andy Jassy Written In a recent LinkedIn post, the company said it has already seen “$260 million in annualized efficiency gains” after integrating its coding assistant, Q, into its workflow.

Microsoft-owned GitHub Copilot is already a big driver of revenue for the tech giant. Satya Nadella The company July financial results announcement Copilot was said to be now “a business bigger than all of Github” when Microsoft acquired it, but investors told me that developers are sufficiently dissatisfied with Copilot that they are open to looking for other solutions.

Tech giants and unicorn startups will undoubtedly own the lion’s share of the coding automation business, but investors say there’s big potential for niche tools that solve specific problems or act as active agents, such as Factory’s “droids” that can automate different parts of the code stack.

There’s still a lot that remains to be proven, and many of the claims about coding ability will likely prove to be premature, or worse. Cognition Labs’ agent product, Devin, was heavily promoted after a demo this spring, with the promise of being the world’s first “AI software engineer.” But it’s unclear if Devin will be able to do it. Selected demos.

Bessemer noted that many coding tasks are inherently well suited to algorithmic assistance. Janelle Teng Because problems must be solved through a clear, fixed set of steps.

“Most modern engineering teams spend so much time on routine tasks like maintaining, securing, refactoring, and testing code that they only have a small amount of time to spend on writing entirely new code,” Teng told me via email. “We see great potential for AI to have a significant impact in these ‘routine’ areas.”

For software developers, coding assistants are kind of friends and foes. On one hand, they are extremely useful for the job, but on the other, they are also very rude. a potential threat to their jobsTeng doesn’t think experienced developers will reach that standard anytime soon, and the profession Split Whether AI assistants can independently write code to developer-grade standards, and issues like cybersecurity and copyright enforcement will require human oversight for years to come.

“We have no intention of eliminating the field, but for people who graduated from coding bootcamps and got jobs making $200,000 a year, it’s going to be harder to find that work,” one investor told me privately.

Greylock Investors Colin Reilly Detailed Market Map It’s a space we’ve been using to keep track of various coding assistant startups and their use cases. Check it out below.

Source: “Work smarter, not harder” Greyrock
Large chart

#coding #funding #frenzy…down #rounds #goodness #declining…Miami #loses #luster

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top