From left to right: BI's Ben Bergman, Mercury's Dan Kang, Cribl's Zach Johnson and Vercel's Marten Abrahamsen. Photo Courtesy of CRV, Tyler Mussetter 2025-05-26T16:00:01Z Save Saved Read in app This story is available exclusively to Business Insider subscribers. Become an Insider and start reading now. Have an account? CFOs of late-stage tech startups face challenges amid market uncertainty and IPO jitters. Mercury, Vercel, and Cribl's CFOs spoke on a panel this week at VC firm CRV. Executives are hopeful that advances in AI can yield returns later this year. With a shaky IPO market, tariff uncertainty, and stock market jitters, these are not easy times to be the chief financial officer of a late-stage tech company.Against that precarious backdrop, I sat down last week with the CFOs of Mercury, Vercel, and Cribl at the San Francisco office of CRV, one of Silicon Valley's oldest venture firms and an early investor in all three startups."I'm expecting a lot more uncertainty," said Daniel Kang, CFO of Mercury, a fintech banking startup that recently doubled its valuation to $3.5 billion after raising $300 million in its latest funding round. "There's a lot of impact from what's happening in DC."All the turmoil means CFOs have to be more nimble, said Kang.Marten Abrahamsen, Vercel's CFO, was more upbeat. He does not expect a recession this year and predicts a stock market rally in the fall."I think a lot of this is going to be fueled by some of the investments we see in AI, and we're already seeing it for some of our products that weren't even here a year ago," said Abrahamsen. "I'm very, very bullish on the remainder of this year and beyond."After President Donald Trump announced sweeping tariffs on imports from other countries on April 2, investors panicked and companies from the payments lender Klarna to the physical therapy startup Hinge Health halted their IPO plans.The pause turned out to be short-lived.Markets have rebounded after Trump rolled back the most severe tariffs and he said he would not fire Federal Reserve Chair Jerome Powell. Bankers are telling companies to go public while the window is open.This week, Hinge Health shares jumped 17% in its market debut after eToro, an Israeli trading platform, made a successful public debut on the Nasdaq, opening 34% above its IPO price. (Klarna's IPO is still on hold after the company reported mounting losses.)Abrahamsen does not think companies should wait until a better market comes along to IPO; instead, they should focus on what they can control."There has been a fear of going public in Silicon Valley," he said. "Great companies can go public even if there's not a hot market out there. If you're an outstanding business, there's always going to be an opportunity."Asked why so few companies are going public, the panelists said companies do not want to deal with the headaches of being a public company when there is so much private financing available. There is also little pressure to IPO from investors and employees, according to Zachary Johnson, CFO of Cribl, a data management solutions startup that raised $319 million last year at a $3.5 billion valuation."They understand that we're trying to build something that's going to be generational," said Johnson. "When we think about how we want to build this company, it's really about focusing on that durability and sustainability of growth."Johnson is hopeful that advances in AI can make Cribl even more attractive to investors when it goes public. He recently tasked everyone on his executive team to come up with an AI initiative."There's some work to be done, but I'm optimistic that we can actually get some real returns on that by the end of this year," he said. "We're still in the early innings of AI."

The team behind Vercel’s V0, an AI-powered platform for web creation, has developed an AI model it claims excels at certain website development tasks. Available through an API, the model, called “v0-1.0-md,” can be prompted with text or images, and was “optimized for front-end and full-stack web development,” the Vercel team says. Currently in beta, it requires a V0 Premium plan ($20 per month) or Team plan ($30 per user per month) with usage-based billing enabled. The launch of V0’s model comes as more developers and companies look to adopt AI-powered tools for programming. According to a Stack Overflow survey last year, around 82% of developers reported that they’re using AI tools for writing code. Meanwhile, a quarter of startups in Y Combinator’s W25 batch have 95% of their codebases generated by AI, per YC managing partner Jared Friedman. Vercel’s model can “auto-fix” common coding issues, the Vercel team says, and it’s compatible with tools and SDKs that support OpenAI’s API format. Evaluated on web development frameworks like Next.js, the model can ingest up to 128,000 tokens in one go. Tokens are the raw bits of data that AI models work with, with a million tokens being equivalent to about 750,000 words (roughly 163,000 words longer than “War and Peace”). Vercel isn’t the only outfit developing tailored models for programming, it should be noted. Last month, JetBrains, the company behind a range of popular app development tools, debuted its first “open” AI coding model. Last week, Windsurf released a family of programming-focused models dubbed SWE-1. And just yesterday, Mistral unveiled a model, Devstral, tuned for particular developer tasks. Companies may be keen to develop — and embrace — AI-powered coding assistants, but models still struggle to produce quality software. Code-generating AI tends to introduce security vulnerabilities and errors, owing to weaknesses in areas like the ability to understand programming logic. Techcrunch event Join us at TechCrunch Sessions: AI Secure your spot for our leading AI industry event with speakers from OpenAI, Anthropic, and Cohere. For a limited time, tickets are just $292 for an entire day of expert talks, workshops, and potent networking. Exhibit at TechCrunch Sessions: AI Secure your spot at TC Sessions: AI and show 1,200+ decision-makers what you’ve built — without the big spend. Available through May 9 or while tables last. Berkeley, CA | June 5 REGISTER NOW

When you purchase through links on our site, we may earn an affiliate commission. Here’s how it works. Vercel releases first AI model for v0, now in beta David Uzondu Neowin · May 22, 2025 02:18 EDT Google recently showered us with AI goodies, including Gemma 3n, an AI model that's designed to run on low-end devices, like smartphones. Now, Vercel has stepped further into the ring with its own generative UI system, v0, by releasing its very first dedicated model. If you do not know what v0 is, it is a sort of competitor to tools like the recently announced Google Stitch, which also aims to let you describe a user interface and have AI generate the design. The tool first saw the light of day back in 2023 as an invite-only beta, promising to turn natural language into front-end code. The newly available model is dubbed v0-1.0-md, and Vercel states it is specifically designed for building modern web applications. This multimodal model supports both text and image inputs, offers a 128,000-token context window with a 32,000-token output limit, and is priced at $3 per million input tokens and $15 per million output tokens. It offers features like 'auto-fix' for common coding blunders and 'quick edit' for streaming inline changes as they are generated. Crucially, v0-1.0-md uses an OpenAI-compatible API, meaning you can plug it into existing tools like Cursor, Codex, or your own custom applications that already speak OpenAI's language, including Vercel's own AI SDK. It even supports function and tool calls, and promises low-latency streaming responses. Developers can poke around with this new model in the Vercel AI Playground to see how it handles different prompts. Currently, access to the v0 API, and thus the v0-1.0-md model, is in beta, and you will need a Premium or Team plan on Vercel with usage-based billing enabled. To get started, you would grab an API key from v0.dev and then send requests to its POST api.v0.dev/v1/chat/completions endpoint, authenticating with a bearer token. While there are daily message limits around 200 messages and context size constraints that mirror its advertised capabilities, Vercel notes you can request higher limits if you hit those ceilings. If you want to dig into the details or see how to set it up, the official v0 docs on Vercel's site have everything you need, including examples. Tags Report a problem with article Follow @NeowinFeed

What the new wave of new AI design and dev tools, — Bolt, V0, Lovable, and Figma Make — mean for the future of software design.Prompt by the author, image generated by Sora.This article builds on reflections I shared last July in The expanded scope and blurring boundaries of AI-powered design, outlining what’s changed in a short time, and what it means for those designing software and leading design teams.Like many others, I’ve been exploring tools like Bolt, Lovable, V0, and most recently Figma Make, looking at how they are changing the way we build software today, and what that means for the future. For those who may not know, these tools are part of a new wave of AI-powered design and development platforms that aim to speed up how we go from prompt to prototype, automating front-end code, generating UI from prompts, and bridging the gap between design and engineering. Bolt is now the second fastest-growing product in history, just behind ChatGPT.While the AI hype hasn’t slowed since ChatGPT’s launch, it’s quickly becoming apparent that these tools represent a step change, one that is rapidly reshaping how we work, and how software gets built.A example of the Bolt.new UI interfaceThis shift didn’t start with AIEven before the recent explosion of AI tooling, design teams have been evolving their approach and expanding their scope of impact. Products like Figma enabled more fluid communication and cross-disciplinary collaboration, while design systems and front-end frameworks like Material, Tailwind, Radix and other libraries helped codify and systematise best practices for visual design, interaction an accessibility.This enabled designers to spend more time thinking about the broader systems, increasing iteration cycles — and less time debating padding. While such tools and frameworks helped to elevate the baseline user experience for many products, in enterprise SaaS in particular, they have had their share of criticism from the resulting sea of sameness that they generated. AI tools are now accelerating and amplifying some of the consequences, both positive and negative. These products represent not just a tooling upgrade, but a shift in what design is, who does it, and how teams are built.Design has evolved from the design of objects, both physical and immaterial, to the design of systems, to the design of complex adaptive systems. The evolution is shifting the role of designers; they are no longer the central planner but rather participants within the systems they exist in. This is a fundamental shift — one that requires a new set of values— Joi Ito, MIT Media Lab (Jan 2016)What AI tools are making possibleThis new wave of AI tools can generate high-quality UIs from a prompt, screenshot, or Figma frame. Work that once required a multidisciplinary team and weeks of effort — from concept to coded prototype — can now happen in a matter of hours. Best practices are baked in. Layouts are responsive by default. Interaction logic is defined in a sentence. Even connecting to real data is no longer a blocker, it’s part of the flow.Lovable, one of the many new AI design and full-stack development tools launched recentlyThese tools differ from popular IDE-based assistants like Cursor, Copilot and Windsurf in both purpose and level of abstraction. UI-based tools like Bolt automate many of the more complex and often intimidating parts of the developer workflows; spinning up environments, scaffolding projects, managing dependencies, and deploying apps. That makes sense, given that many of them were built by hosting platforms such as Vercel and Replit.With this new speed and ease of use, designers don’t need to wait on engineers to see how something feels in practice. They can test ideas with higher fidelity faster, explore more variations, and evolve the experience in tight feedback loops.Figma Make: Start with a design and prompt your way to a functional prototype, fast — all in Figma.This shift has also given rise to what some are calling ‘Vibe coding’, a term coined by Andrej Karpathy, that captures this expressive, real-time way of building software. Instead of following a strict spec or writing code line by line, you start with a vibe or loose concept, and use these tools to sculpt the idea into something functional. You prompt, tweak, adjust components, and refine until it feels right. It’s intuitive, fast, and fluid.There’s a new kind of coding I call “vibe coding”, where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. It’s possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good.— Andrej KarpathyIn this new paradigm, the output isn’t just faster, it’s driven by rapid judgement and intuition, not necessarily depth of technical experience. In addition, the barrier to entry for non-designers to explore ideas has lowered too. Now that anyone can create compelling, usable apps with front-end and back-end logic, what does that mean for design?I would love to say that this means more time spent on outcomes and higher-impact work for designers, but it’s more likely to disrupt the foundations of what it means to be a designer. The boundaries between the classic product triad of design, engineering and product management were already blurring but this looks like it will accelerate even more.We are in the middle of a significant industry shift, we’re heading into a period of rapid, unpredictable change.. While testing some of these new AI tools, I have had several ‘oh shit’ moments where I get a sense of how things might evolve…. this is what copywriters and others in similar writing roles must have felt when ChatGPT first came out.The author, while vibe coding (image via Giphy)What this might mean for designAs UI generation becomes commoditized, the value of design shifts upstream. With that, the scope of what is expected from design will shift. Future designs team are likely to be smaller, and more embedded in product strategy. As companies grow, design functions won’t necessarily need bigger design teams, they will need higher-leverage ones.Meanwhile, designers and engineers will work more closely together — not through handoff, but through shared tools and live collaboration. In enterprise environments in particular, much of the engineering work is not so much about zero-to-one implementation but about working within and around established technical constraints. As front-end becomes commoditized, engineers will shift their focus further upstream to establishing strong technical foundations and systems for teams to build from.From years of experience to mindsetSome worry this shift will reduce opportunities for junior designers. It’s true there may be fewer entry-level roles focused on production work. But AI-native designers entering the field now may have an edge over seasoned professionals who are tied to traditional methods.In an AI-driven world, knowing the “right” design process won’t matter as much. Technical skills, domain expertise and a strong craft will still help, but what really counts is getting results — regardless of how you get there.The greatest danger in times of turbulence is not the turbulence, it is to act with yesterday’s logic— Peter DruckerMindset will matter more than experience. Those who adapt fast and use AI to learn new domains quickly will stand out. We are already starting to see this unfold. Tobi Lutke, CEO of Shopify recently stated that AI usage is now a baseline expectation at Shopfiy. He went even further, starting that “Before asking for more headcount and resources, teams must demonstrate why they cannot get what they want done using AI”.This demonstrates that adaptability and AI fluency are becoming core expectations. In this new landscape, titles and years of experience will matter less. Designers who can leverage AI as a force multiplier will outpace and outshine those relying on traditional workflows or rigid processes.Speed isn’t everythingNote that I didn’t use the word taste, which many now describe as critical in the AI era. I cringe a little when I hear it — taste feels vague and subjective, often tied to the ‘I’ll know it when I see it’ mindset the design industry has been trying to shake off for years. While I get the intent, I prefer to describe this as judgment: the ability to make calls informed by experience and grounded in clear intent, shared principles, and a solid grasp of user and technical context — not personal preference or aesthetic instinct. When you can create infinite variations almost instantly, judgment is what helps you identify what’s truly distinct, useful and worth refining.What does this mean for designing within enterprise environmentsI lead the design team at DataRobot, a platform that helps AI builders create and manage agentic, generative and predictive workflows within large enterprises. We’ve been exploring how AI tools can augment design and development across the org.Screens from the DataRobot AI platformWhile these tools are great for initial ideation, this is often only a small part of the work in enterprise environments. Here, the reality is more complex: teams work within deeply established workflows, technical frameworks, and products with large surface areas.This differs from consumer design, where teams often have more freedom to invent patterns and push visual boundaries. Enterprise design is about reliability, scalability, and trust. It means navigating legacy systems, aligning with highly technical stakeholders, and ensuring consistency across a broad suite of tools.For us, one of the clearest use cases for AI tooling has been accelerating early-stage concepting and customer validation. While most of our focus is on providing infrastructure to build and manage AI models, we’ve recently expanded into custom AI apps, tailored for specialized workflows across a broad range of industries and verticals. The number of UI variants we would need to support is simply too vast for traditional design processes to cover.Some examples of DataRobot applications — both production and concept.In the past, this would have meant manually designing multiple static iterations and getting feedback based on static mocks. Now, AI tools let us spin up tailored interfaces, with dynamic UI elements tailored for different industries and customer contexts, while adhering to our design system and following best practices for accessibility. Customers get to try something close to the real output and we get better signal earlier in the cycle, reducing wasted effort and resources.In this context, the strict frameworks used by tools like V0 (like Tailwind) are an advantage. They provide guardrails, meaning you need to go out of your way to create a bad experience. It’s early days, but this is helping non-designers in particular to get early-stage validation with customers and prospects.This means the role of the design team is to provide the framework for others to execute, creating prompt guides that codify our design system and visual language, so that outputs remain on brand. Then we step in deeper after direction is validated. In effect, we’re shifting from execution to enablement. Design is being democratized. That’s a good thing, as long as we set the frame.Beyond the baselineAI has raised the baseline. That helps with speed and early validation, but it won’t help you break new ground. Generative tools are by their nature derivative.When everything trends toward average, we need new ways to raise the ceiling. Leadership in this context means knowing when to push beyond the baseline, shaping a distinct point of view grounded in reality and underpinned by strong principles. That point of view should be shaped through deep cross-functional collaboration, with a clear understanding of strategy, user needs, and the broader market.In a world where AI makes it easier than ever to build software, design is becoming more essential and more powerful. It’s craft, quality, and point of view that makes a product stand out and be loved.— Dylan FieldWhat to focus on nowFor individual contributors or those just starting out, it can feel daunting and difficult to know where to start:Start experimenting: Don’t wait for the perfect course, permission or excuse. Just jump in and run small tests. See how you can replicate previous briefs (or current briefs in parallel) in order to get a feel for where they excel and where they break.Look for leverage: Don’t just use these tools to move faster — use them to think differently. How might you explore more directions, test ideas earlier, or involve others upstream?Contribute to the system: Consider how you might codify what works to improve patterns, prompts, or workflows. This is increasingly where high-impact work will live.If you’re leading a design team:Design the system, not just the UI: Build the tools, patterns, and prompts that others can use to move fast.Codify best practices: Think how you might translate tribal knowledge into actionable context and principles, for both internal teams and AI systems.Exercise (your) judgement: Train your team to recognize good from average in the context of your product. Establish a shared language for what good means in your context, and how you might elevate your baseline.Final thoughtsThe UI layer is becoming automated. That doesn’t make design less important — it makes it more critical. Now everyone can ship something decent, but only a great team can ship something exceptional.AI might handle the pixels, but it’s just a tool. Design’s purpose is clearer than ever: understanding users, shaping systems, and delivering better outcomes. AI tools should amplify our capabilities, not make us complacent. This means that while we integrate them into our workflows, we must continue to sharpen our core skills. What Paul Graham said about writing applies equally to design.When you lose the ability to write, you also lose some of your ability to think— Paul GrahamThis article was written with the assistance of ChatGPT 4o.John Moriarty leads the design team at DataRobot, an enterprise AI platform that helps AI practitioners to build, govern and operate predictive and generative AI models. Before this, he worked in Accenture, HMH and Design Partners.Design in the age of vibes was originally published in UX Collective on Medium, where people are continuing the conversation by highlighting and responding to this story.

AI startup Cohere has acquired Ottogrid, a Vancouver-based platform that develops enterprise tools for automating certain kinds of high-level market research. Sully Omar, one of the founders of Ottogrid, announced the deal Friday in a post on X. He didn’t disclose the terms. Ottogrid will sunset its product, according to Omar, but will give customers “ample notice” and “a reasonable transition period.” “We’re very excited to join the Cohere team and integrate Ottogrid into Cohere’s … platform,” Omar said in a statement. “Through our work with Cohere, we’re [going to] dramatically impact how people can automate their workflows, enrich their data, and scale their operations.” Cohere didn’t immediately respond to a request for comment. Cohere’s purchase of Ottogrid comes as the former experiences a bit of corporate turbulence. According to The Information, Cohere fell well short of revenue projections the company prepared in early 2023, missing its target for last year by 85%. The company told Reuters on Thursday that its annualized revenue recently reached $100 million, following a strategic shift with a focus on private AI deployments for customers in sectors like healthcare, government, and finance. Techcrunch event Join us at TechCrunch Sessions: AI Secure your spot for our leading AI industry event with speakers from OpenAI, Anthropic, and Cohere. For a limited time, tickets are just $292 for an entire day of expert talks, workshops, and potent networking. Exhibit at TechCrunch Sessions: AI Secure your spot at TC Sessions: AI and show 1,200+ decision-makers what you’ve built — without the big spend. Available through May 9 or while tables last. Berkeley, CA | June 5 REGISTER NOW Ottogrid launched in 2023 as Cognosys, led by Omar and Homam Malkawi. It rebranded in October 2024 with a major platform redesign that introduced a number of new integrations, tools, and APIs. Today, Ottogrid offers a “native table interface” with AI-powered document analysis capabilities. Customers can use it to extract data from a website and save it directly to a spreadsheet, for example, or automatically enrich sales lead lists. Ottogrid managed to raise $2 million in venture capital from investors, including GV (Google Ventures), Untapped Capital, Replit CEO Amjad Masad, Vercel CEO Guillermo Rauch, Cohere co-founders Ivan Zhang and Aidan Gomez, and others prior to its exit, according to Crunchbase. As part of Cohere, Omar says that Ottogrid will focus primarily on North, Cohere’s recently launched ChatGPT-style application designed to assist knowledge workers with tasks such as summarizing documents.

A new global phishing threat called "Meta Mirage" has been uncovered, targeting businesses using Meta's Business Suite. This campaign specifically aims at hijacking high-value accounts, including those managing advertising and official brand pages. Cybersecurity researchers at CTM360 revealed that attackers behind Meta Mirage impersonate official Meta communications, tricking users into handing over sensitive details like passwords and security codes (OTP). The scale of this operation is alarming. Researchers have already identified over 14,000 malicious URLs, a concerning majority of which—nearly 78%—were not blocked by browsers at the time the report was published. Cybercriminals cleverly hosted fake pages leveraging trusted cloud platforms like GitHub, Firebase, and Vercel, making it harder to spot the scams. This method aligns closely with recent findings from Microsoft, which highlighted similar abuse of cloud hosting services to compromise Kubernetes applications, emphasizing how attackers frequently leverage trusted platforms to evade detection. The attackers deploy fake alerts about policy violations, account suspensions, or urgent verification notices. These messages, sent via email and direct messages, look convincing because they mimic official communications from Meta, often appearing urgent and authoritative. This tactic mirrors techniques observed in the recent Google Sites phishing campaign, which used authentic-looking Google-hosted pages to deceive users. Two main methods are being used: Credential Theft: Victims enter passwords and OTPs into realistic-looking fake websites. The attackers deliberately trigger fake error messages, causing users to re-enter their details, ensuring accurate and usable stolen information. Cookie Theft: Scammers also steal browser cookies, allowing them continued access to compromised accounts even without passwords. These compromised accounts don't just affect individual businesses—they're often exploited to run malicious advertising campaigns, further amplifying damage, similar to tactics observed in the PlayPraetor malware campaign that hijacked social media accounts for fraudulent ad distribution. CTM360's report also outlines a structured and calculated approach used by the attackers to maximize effectiveness. Victims are initially contacted with mild, non-alarming notifications that progressively escalate in urgency and severity. Initial notices might mention generic policy violations, while subsequent messages warn of immediate suspensions or permanent deletion of accounts. This incremental escalation induces anxiety and urgency, driving users to act quickly without thoroughly verifying the authenticity of these messages. To protect against this threat, CTM360 recommends: Only use official devices to manage business social media accounts. Use separate business-only email addresses. Enable Two-Factor Authentication (2FA). Regularly review account security settings and active sessions. Train staff to recognize and report suspicious messages. This widespread phishing campaign underscores the importance of vigilance and proactive security measures to protect valuable online assets. Found this article interesting? This article is a contributed piece from one of our valued partners. Follow us on Twitter  and LinkedIn to read more exclusive content we post.