If there’s one thing that VCs agree on when backing AI startups, it’s that AI requires a different investment approach than prior technological shifts.

“It’s a funky time,” said Aileen Lee, founder and managing partner of Cowboy Ventures, on stage at TechCrunch Disrupt 2025. The longtime VC noted that the rules of investing have significantly shifted now that some AI companies are leaping from “zero to $100 million in revenue in a single year.”

However, Lee also noted that, based on her firm’s research, Series A investors aren’t just seeking rapid revenue growth. “It’s an algorithm with different variables and different coefficients.”

Some of the factors investors now measure, according to Lee, include whether the startup is generating data, the strength of its competitive moat, the founders’ past accomplishments, and the technical depth of the product. “Depending on what your company is, the output of the algorithmic formula is going to be different,” she said.

Jon McNeill, co-founder and CEO of startup creation firm DVx Ventures, stated that even startups that grow rapidly from inception to $5 million in revenue often struggle to secure follow-on funding. “I think this game has changed, and it is changing dynamically,” he said.

McNeill noted that Series A investors are now applying the same rigorous standards to seed-stage startups that they previously reserved for more mature companies.

“I think a lot of investors have figured out that the breakout companies, in most cases, don’t have the best tech,” McNeill said, about why Series A VCs are looking so closely at startups’ ability to attract and retain customers. “They have the best go-to market.”

Steve Jang, founder and managing partner of Kindred Ventures, disagreed that a strong go-to-market (GTM), an industry term for sales and marketing, holds greater weight for investors. “I don’t think it’s 100% true to say mediocre technology, great GTM wins and raises money and gets customers. I think that it’s a necessary requirement to have both.”

While McNeill later clarified that having a solid product is important, he indicated that his initial comment was related to the founders’ need to develop an exceptionally strong sales and marketing strategy right out of the gate. “Investors are getting much more sophisticated on the go-to market, than they have in the past,” he said.

(The debate over marketing versus tech was brought to the forefront later during the conference when Roy Lee, founder of the viral startup Cluely, said on stage that launching a product that barely worked, even with massive social media fame, may not always be the best idea.)

Lee added that AI startups are now also under pressure to deliver product updates and new features at an unprecedented pace, preempting existing companies that might try to introduce similar products.  “If you look at how much OpenAI and Anthropic are shipping, you’re going to have to figure out how to match how much you ship, how quickly and the quality of it,” she said.

Despite the expectations for breakneck growth and fast product development, panelists agreed that the AI industry is still in its very early stages. As Jang put it, “There are no clear, outright winners, even in LLMs. There are competitors nipping at their heels.”

This means startups still have a path to unseating perceived leaders, whether they are decades-old companies or fast-moving newcomers.

Music streaming service Spotify said it is introducing a new shuffle option that repeats fewer suggested songs.

The company originally launched Smart Shuffle for paid users in 2023, a feature that added new songs to a playlist’s queue based on users’ previously played songs and their preferences. Users previously could not disable this feature.

With this new update, the company is making shuffle with fewer repeats the default for paid users. This means users will hear fewer songs from their recent listening history and more new songs.

If they want the standard Smart Shuffle, they can go to Settings > Playback and choose the “Standard” option under the Shuffle Mode section.

Spotify added that both free and premium users can tap any song to play next while shuffling without rearranging the queue.

Lauren Saunders, product director for personalization at Spotify, said that shuffle follows randomness, which means there is a higher probability of listening to the same songs again.

“That’s the thing about real randomness: It can be clumpy. Just like you can roll three sixes in a row, a purely random shuffle might never seem to play the song you’re hoping for, or it might stack the same artist or album closer together than your brain expects. The math is right, but the feeling is wrong,” Saunders said.

“Instead of giving you one random order and calling it a day, we now generate hundreds of truly random versions of your playlist. Then we score each one for freshness, looking at how recently you’ve played certain songs, how much variety is packed into the opening stretch, and whether you’re getting repeats too soon,” she said about the new approach to shuffle.

Along with this feature, Spotify is also releasing an AI-powered recap feature for audiobooks to let users catch up on a story to the point where they stopped listening. Users will see a recap button at the top of the audiobook’s page, which will play a summary of the story so far.

Spotify said that the recaps will show up after users have listened to at least 15-20 minutes of a book and are regularly updated as they progress. It added that Spotify is not using audiobook content to train models, and recaps don’t replicate the original narration.

The feature is currently available on iOS in beta for a selection of English-language titles.