Battlelines are being drawn between the major AI labs and the popular applications that rely on them.<\/p>\n<\/div>\n
This week, both Anthropic and OpenAI took shots at two leading AI apps: Windsurf, one of the most popular vibe coding tools, and Granola, a buzzy AI app for taking meeting notes.<\/p>\n<\/div>\n
\u201dWith less than five days of notice, Anthropic decided to cut off nearly all of our first-party capacity to all Claude 3.x models,\u201d Windsurf CEO Varun Mohan<\/strong> wrote on X<\/a> this week, noting that \u201cwe wanted to pay them for the full capacity.\u201d An additional statement on Windsurf\u2019s website<\/a> said: \u201cWe are concerned that Anthropic\u2019s conduct will harm many in the industry, not just Windsurf.\u201d<\/p>\n<\/div>\n Here, Mohan\u2019s company is collateral damage in Anthropic\u2019s rivalry with OpenAI, which has reportedly been in talks to acquire Windsurf for about $3 billion. The deal hasn\u2019t been confirmed, but even the spectre of it happening was enough for Anthropic to cut off one of the most popular apps that it powers. After a spokesperson told TechCrunch\u2019s<\/em> Maxwell Zeff<\/strong><\/a> that Anthropic was \u201cprioritizing capacity for sustainable partnerships,\u201d co-founder Jared Kaplan<\/strong> put it more bluntly. <\/p>\n<\/div>\n \u201cWe really are just trying to enable our customers who are going to sustainably be working with us in the future,\u201d Kaplan told Zeff<\/a>. \u201cI think it would be odd for us to be selling Claude to OpenAI.\u201d<\/p>\n<\/div>\n Meanwhile, OpenAI sent its own warning shot this week to the budding AI app ecosystem. It announced a \u201crecord mode\u201d<\/a> for ChatGPT \u2014 initially only for enterprise accounts \u2014 that transcribes calls and generates meeting notes. This is the core use case of Granola, one of my favorite AI tools that recently raised $43 million<\/a> in additional funding and released a mobile app. <\/p>\n<\/div>\n Given how quickly Granola has evolved to do more<\/a> than summarize meetings, I suspect that the company isn\u2019t at risk of extinction. Still, it will be harder to grow when hundreds of millions of ChatGPT users eventually have access to its main functionality.<\/p>\n<\/div>\n It\u2019s unclear how the tension between the product ambitions of OpenAI and Anthropic and the needs of their API customers will settle out. When I interviewed Anthropic\u2019s chief product officer, Mike Krieger<\/strong>, back in March<\/a>, the company had just announced its own Claude coding competitor to Windsurf and Cursor, which coincidentally raised $900 million<\/a> this week. I asked Krieger the obvious question: how does Anthropic think about competing with its API customers? He didn\u2019t really have an answer. <\/p>\n<\/div>\n \u201cI think this is a really delicate question for all of the labs and one that I\u2019m trying to approach really thoughtfully,\u201d Krieger told me at the time. \u201cHopefully, we\u2019ll all be able to navigate the occasionally closer adjacencies.\u201d<\/p>\n<\/div>\n AI investor Zak Kukoff <\/strong>put it<\/a> well this week: \u201cAt some point model providers are going to need to decide if they want to be stable platforms or compete for every vertical.\u201d<\/p>\n<\/div>\n Ultimately, this week served as a wake-up call for the many startups building businesses on the backs of AI models; if you are successful enough, you run the risk of being copied by your model provider. A lot of companies are thinking through this risk right now, especially as OpenAI builds a new team<\/a> to help its API customers \u201ctranslate abstract ideas into production applications.\u201d<\/p>\n<\/div>\n \u201cYou have to wonder if the recent moves by the big AI labs to more directly compete with the app layer will be one giant tailwind for incumbents like Google, Amazon, MSFT, etc.,\u201d Michael Mignano<\/strong>, a Granola board member, wrote this week<\/a>. \u201cIf developers can\u2019t trust the labs, maybe it\u2019s better to trust the big guys like they did for cloud?\u201d<\/p>\n<\/div>\n This week, I heard two CEOs contradict the growing fear<\/a> that AI will destroy jobs en masse, at least when it comes to engineering roles. <\/p>\n<\/div>\n The first was Sundar Pichai<\/strong>, whom I watched speak at Bloomberg\u2019s tech conference in San Francisco. He downplayed Dario Amodei\u2019s <\/strong>doomerism fear about job loss, correctly pointing out that \u201cwe\u2019ve made predictions like that for the last 20 years about technology and automation, and it hasn\u2019t quite played out that way.\u201d He went so far as to say, \u201cI expect we will grow from our current engineering base into next year,\u201d because AI \u201callows us to do more.\u201d<\/p>\n<\/div>\n The next day, I walked down the street to the Moscone Center to see Snowflake CEO Sridhar Ramaswamy<\/strong>, who had just spoken to a room of 4,000 developers with AI pioneer Andrew Ng<\/strong>. I asked Ramaswamy if AI had changed his hiring plans, and he said he agreed with a ranking of hiring desirability for engineers that Ng had just described onstage, with the top being experienced engineers who leverage AI tools, followed by early-career engineers who are all-in on AI. He noted that new graduates who avoid AI tools are at the bottom of the desirability ranking and may struggle to find jobs.<\/p>\n<\/div>\n If anything, it\u2019s the middle of the workforce \u2014 those who are in the middle of their careers and hesitant to adopt AI tools \u2014 that is the most in danger of near-term displacement, Ramaswamy argued. \u201cCompanies tend to accrete middle management, so there\u2019s very much a push to get more people who are doing<\/em>. How do we get them as leveraged as possible? Snowflake has historically been a little top-heavy on the engineering side, so we are balancing that out.\u201d<\/p>\n<\/div>\n \u201cOh, man, the girls are fighting, aren\u2019t they?\u201d – Rep. Alexandria Ocasio-Cortez<\/strong> commenting on<\/a> what was the best day on Twitter in years.<\/p>\n<\/div>\n \u201cMaybe there\u2019s a world where you have one AI in the sky. Maybe you actually have a bunch of domain-specific agents that require a bunch of specific work to make it happen. I think the evidence has really been shifting towards this menagerie of different models.\u201d – OpenAI\u2019s Greg Brockman<\/strong> speaking at<\/a> the AI Engineer\u2019s World Fair. <\/p>\n<\/div>\n \u201cGive it a year. We\u2019ll be doing a billion queries a week if we can sustain this growth rate.\u201d – Perplexity CEO Aravind Srinivas<\/strong> onstage at Bloomberg\u2019s tech conference<\/a>.<\/p>\n<\/div>\n \u201cWe were accidentally cash flow positive in Q1, which was cool.\u201d – Substack CEO Chris Best<\/strong> speaking at<\/a> The Information\u2019s creator economy summit. <\/p>\n<\/div>\n If you haven\u2019t already, don\u2019t forget to subscribe to The Verge<\/em><\/a>, which includes unlimited access to Command Line<\/em><\/a> and all of our reporting.<\/p>\n<\/div>\n As always, I welcome your feedback, especially if you\u2019ll be attending WWDC next week as well, or if you have a story idea to share. You can respond here or ping me securely on Signal<\/a>.<\/p>\n<\/div>\nA different take on AI and job loss<\/strong><\/h2>\n<\/p>\n<\/div>\n
\n