February 2026 Summaries

Gamma is revolutionizing presentation design by automating the creation process, allowing users to generate well-structured presentations with a single sentence while maintaining focus on content. Initially, Gamma's AI generated slides from prompts, but with the introduction of Gamma Agent in October 2025, the company shifted towards collaborative workflows, enabling users to iteratively refine presentations through conversational editing. This transition necessitated sophisticated conversation management, prompting the use of a lower-level API within the AI SDK for enhanced control over conversation state. Gamma's model-agnostic architecture enables rapid integration of new image models, facilitating the development of new features with minimal configuration. The company leverages Vercel's deployment tools for efficient and reliable updates, maintaining an impressive deployment rate despite a small engineering team. Gamma's serverless functions ensure fast, responsive editing and presentation loading, highlighting its commitment to seamless user experiences. As agents become more advanced, Gamma emphasizes the importance of context in enhancing agent utility, focusing on integrating brand guidelines and user history into its models to deliver tailored, impactful communication solutions.

Avalara, a company that connects businesses to over 1,400 systems for automating tax compliance globally, has transformed its product development process by leveraging Vercel's v0 AI-native platform. This new approach, spearheaded by Chief Strategy and Product Officer Jayme Fishman, focuses on modernizing Avalara's building methods through digital transformation and AI innovation. Vercel's v0 allows Avalara to quickly translate plain language ideas into working prototypes, significantly reducing the time from concept to execution. This shift has enabled the creation of two patent-pending products in just 60 days, reshaping Avalara's product and design processes by eliminating lengthy handoffs between teams and fostering real-time collaboration and iteration. The introduction of v0 has not only streamlined Avalara's workflow but also enhanced their ability to support customers more effectively by offering tools like a Chrome extension that integrates seamlessly with users' workflows.

Vercel developed an innovative system called the Community Guardian to streamline and enhance interactions within their developer community by automating routine tasks such as routing, triage, and follow-ups, thereby allowing team members to focus on more meaningful connections and complex problem-solving. The Guardian, which assigns posts to the right team members and ensures that no queries are left unanswered, operates using AI Gateway and Vercel Workflows for efficient resource management. Complementing the Guardian is c0, a research assistant that integrates with Slack to provide team members with comprehensive context packages by scouring through knowledge bases, documentation, and past discussions, ultimately facilitating faster and more accurate responses. Both tools allow Vercel to maintain a balanced workload while boosting operational efficiency and community engagement, and they demonstrate how even non-developers can implement sophisticated automation solutions by clearly defining problems and iteratively refining their approach using AI tools like AI SDK and Vercel Workflows.

OpenEvidence, the fastest-growing clinical decision support platform in the U.S., recently experienced a viral moment on TikTok that tested its ability to handle massive user traffic without compromising reliability. Despite reaching over two million views in a week, the platform maintained fast response times and low error rates, thanks to its hybrid architecture that leverages Python on Google Cloud Platform for backend processes and Next.js on Vercel for frontend operations. This infrastructure allows for seamless scaling and quick deployments, essential for supporting over 20 million clinical consultations monthly and aiding in medical decisions for over 100 million Americans annually. The introduction of Vercel's Fluid compute further reduced serverless costs by 90%, enhancing performance and efficiency as OpenEvidence scaled to 1000x growth. With a small team, the platform continues to meet the high standards of speed and reliability expected in healthcare, proving capable of modernizing medical software while maintaining the stability of traditional systems. OpenEvidence serves over 40% of U.S. physicians, providing a crucial tool for making life-saving clinical decisions grounded in peer-reviewed medical literature.

As agents increasingly adopt coding agent architectures, which involve reading filesystems, running shell commands, and generating code, they become multi-component systems requiring distinct levels of trust. Traditionally, these components operate within a single security context, exposing them to risks such as prompt injections, which can lead to unauthorized actions like data exfiltration. To mitigate these risks, it's crucial to establish security boundaries among the four actors in an agentic system: the agent, agent secrets, generated code execution, and the filesystem. Current practices often lack these boundaries, allowing shared security contexts that heighten vulnerability. A more secure approach involves separating agent compute from sandbox compute, with generated code running in isolated environments without direct access to secrets. Combining application sandboxes with secret injection proxies enhances security by maintaining isolation and preventing credential misuse. Despite the benefits, adoption remains low due to compatibility challenges, but those who implement these measures early will gain a security edge as agents handle increasingly sensitive tasks.

Skills Night, an event for the open skills ecosystem known as skills.sh, highlighted the rapid growth and integration of skills into various coding agents, with over 69,000 skills and 2 million CLI installs to date. Originating from Shu Ding's extensive documentation of React knowledge, the platform has evolved into a package manager for agent context, facilitating the distribution and installation of skills with ease. As the ecosystem expands, security concerns have been addressed through partnerships with Gen, Socket, and Snyk to audit skills for potential vulnerabilities. Demos at the event showcased how skills enhance agents' capabilities by providing context, with successful applications in automating iOS feature upgrades and integrating skills as infrastructure in documentation and code analysis tools. The event underscored the importance of delivering relevant context to improve agent output and efficiency, with ongoing development and community engagement at the heart of its future growth.

AI Gateway has introduced a beta feature for video generation, available to Pro and Enterprise plan users, leveraging AI SDK 6 to create photorealistic videos with synchronized audio through a unified API. This feature supports multiple modes of video creation—text-to-video, image-to-video, first and last frame, and reference-to-video—each providing unique capabilities for generating content from prompts, images, and reference materials, with options for cinematic quality and identity consistency across scenes. The platform offers four initial video models with 17 variations, including Grok Imagine for fast instruction-following, Wan for reference-based generation, Kling for image-to-video transitions, and Veo for high visual fidelity, each suited to different creative needs like ad creation, storytelling, and brand consistency. Users can experiment with these models in the AI Gateway Playground without coding, utilizing the platform's comprehensive dashboard for a streamlined video generation workflow.

The text discusses performance optimization efforts in server-side streaming, focusing on the use of WebStreams in Node.js, particularly in the context of frameworks like Next.js and React. It highlights the inefficiencies found in native WebStreams, primarily due to Promise and object allocation overhead, and introduces a library called fast-webstreams that reimplements the WHATWG Streams API using optimized Node.js streams. The library significantly improves throughput by minimizing Promise creation and leveraging Node.js's C++ internals, achieving substantial performance gains in scenarios common in server-side rendering, such as React Server Component streaming and multi-transform chains. The ultimate aim is to integrate these optimizations into Node.js itself, with contributions already underway, to enhance the performance of WebStreams without the need for external libraries.

Stably, a six-person startup, experienced significant growth and efficiency improvements by utilizing Vercel's platform to eliminate infrastructure anxiety, allowing them to accelerate the development and deployment of their AI testing agents. Initially facing bottlenecks due to the need for infrastructure decisions and deployment concerns, the company shifted its focus from managing complex DevOps to leveraging Vercel's tools, boosting their ability to launch new products quickly and enhancing their autonomous testing capabilities. Co-founders Jinjing Liang and Neil, with backgrounds from Google Chrome and Uber, emphasized reliability and lean operations, relying on AI tools to optimize their processes. Winning Vercel’s AI Accelerator provided validation and opened doors for enterprise contracts, while the experience reshaped their approach to product development and pricing strategy. As a result, Stably transitioned from treating testing as an afterthought to embedding it directly into the developer workflow, leading to increased enterprise traction and plans to open-source more of their core product.

Stably, a six-person startup, has transformed its operations by leveraging Vercel's platform to overcome infrastructure challenges, winning Vercel's AI Accelerator in the process. Initially hindered by infrastructure decisions and deployment anxiety, Stably's team now rapidly ships AI testing agents by integrating directly into developer workflows, eliminating the bottleneck of manual validation. Co-founders Jinjing Liang and Neil, drawing on their extensive backgrounds with Google Chrome and Uber, have prioritized reliability and efficiency, using AI tools and Vercel's features to maintain a lean, revenue-generating operation. The shift to a streamlined infrastructure has allowed Stably to expedite product launches, scale enterprise contracts, and adopt an open-source approach, echoing Vercel's successful strategies. Winning the AI Accelerator not only provided funding but also enhanced brand credibility, facilitating enterprise sales and opening doors to new opportunities. The company's culture has evolved, promoting decisive action and innovation, as they continue to refine their AI-powered testing framework and expand their product offerings.

Ramp, a financial operations platform that integrates bill payments, corporate cards, procurement, and travel to optimize business efficiency, has successfully rebuilt its marketing site using Next.js and Vercel. This migration involved transitioning the most critical and visited pages individually while performing A/B testing to ensure optimal performance based on Core Web Vitals. The site has impressively maintained 100% uptime even during significant traffic spikes, such as those experienced during stunt-marketing events. Additionally, Ramp has implemented the Vercel AI SDK to develop a semantic search feature to enhance accessibility and organization across its varied internal resources.

AI has transformed information discovery, prompting businesses to understand how large language models (LLMs) search for and summarize web content. Vercel is developing an AI Engine Optimization (AEO) system to track how both standard models and coding agents discover, interpret, and reference their content. While standard models are straightforward to track, coding agents, which operate in development environments, present unique challenges, such as execution isolation and observability. Vercel addresses these by using ephemeral Linux MicroVMs that provide isolated environments for agents to run. The system follows a six-step lifecycle, which includes creating a sandbox, installing the agent CLI, injecting credentials, running the agent, capturing its transcript, and tearing down the sandbox. Different agents produce varied transcript formats, necessitating a normalization process to feed data into a unified pipeline. This pipeline captures, parses, enriches, and summarizes data for brand extraction, applicable to both models and agents. Early findings indicate coding agents often perform web searches and produce actionable code recommendations, highlighting the importance of optimizing content for these agents. Vercel plans to open-source its system and expand its coverage to include more agents and prompt types.

The commoditization of AI models has revolutionized the process of building sophisticated software, allowing individuals to quickly create internal tools and agents, but the challenge lies in deploying and operating these systems effectively at scale. Vercel's agent orchestration platform addresses this by providing robust infrastructure that supports the reliable running of agents, offering features like Sandboxes for secure execution, Fluid compute for scalable resource management, and AI Gateway for routing model requests. This shift has changed the build vs. buy equation, as companies can now create customized solutions with immediate ROI, necessitating a strong internal AI stack akin to external product infrastructure. Vercel enables the seamless operation of agents, facilitating complex workflows and ensuring reliability through durable orchestration, while offering observability tools to optimize performance and troubleshoot issues. As the demand for internal applications grows, Vercel's platform ensures that companies can not only build agents but also run them efficiently, promising a higher ROI and a competitive edge in the evolving technological landscape.

Geist Pixel is a newly launched typeface that expands the Geist font family with a bitmap-inspired design built on a strict pixel grid, designed for practical use rather than as a mere visual novelty. It consists of five distinct variants, each maintaining the typographic precision and consistency required for real-world applications in user interfaces and digital products. Geist Pixel is created to function seamlessly with the existing Geist fonts, offering predictable metrics and strong alignment across layouts. It is crafted for modern web environments, supporting multiple languages and stylistic sets, and is already influencing the design language at Vercel. This typeface was meticulously designed, with each glyph manually refined to ensure clarity and personality across different sizes and platforms. The project was a collaborative effort, involving numerous contributors who ensured its design and implementation were executed with care and attention to detail.

The Vercel AI Accelerator is a six-week program designed to support early-stage AI startups by offering over $6 million in credits from platforms like Vercel, AWS, and leading AI services, alongside community support, mentorship, and networking opportunities. The program, which runs from March 2nd to April 16th, invites 40 global teams to participate in intensive development, culminating in a demo day in San Francisco where participants can showcase their products to industry leaders and investors. Participants benefit from weekly sessions with industry leaders, access to exclusive AI content, community connections, and opportunities to receive investment and resources. The program aims to equip startups with the infrastructure and support needed to build scalable AI applications while fostering innovation through credits and resources from partners like Anthropic, OpenAI, and Hugging Face. Vercel's focus on building autonomous AI agents is reflected in its recent AI releases and its support of startups such as Stably AI and Bear AI, highlighting the program's commitment to aiding startups in solving real-world problems through AI technology. Applications are open until February 16th, with selections based on submission quality, founder background, and potential impact and scalability.

Stably, a six-person startup, tackled infrastructure anxiety by leveraging Vercel's platform, which streamlined their operations and allowed them to focus on their AI-powered end-to-end testing product. This testing framework integrates directly with coding agents, enabling them to autonomously validate their work, thus eliminating the bottleneck of manual testing. Adopting Vercel's infrastructure facilitated rapid deployment and development, transforming the team's approach from cautious to proactive, where new product lines and internal tools could be launched quickly without deployment concerns. Winning Vercel's AI Accelerator provided not only validation and funding but also strategic insights, particularly in pricing, and opened enterprise sales opportunities. This partnership has empowered Stably to convert enterprise pilots into contracts, embrace open-source principles, and maintain a lean yet revenue-generating operation.

Since its general availability in 2024, v0 has empowered over 4 million users to swiftly transform ideas into applications, evolving from a novelty to a critical business tool. The platform addresses the challenges posed by vibe coding, such as security vulnerabilities and inefficiencies in traditional software development processes, by introducing enterprise-grade security, seamless GitHub integration, and a sandbox-based runtime for working with existing codebases. v0 democratizes software development, allowing non-engineers to participate in production-ready coding through enhanced Git workflows, and offers secure data connections for building custom reports. The platform's enhancements enable various teams, including marketers and data analysts, to ship updates and features efficiently, minimizing the traditional roadblocks in the development cycle. As v0 continues to evolve, future updates will focus on enabling end-to-end agentic workflows using AI models, further enhancing its capabilities for production applications.

Agents that browse the web require clean, structured text rather than the CSS, client-side JavaScript, and images that humans typically view, as these elements fill up their context window and consume tokens without providing useful information. To address this, a new update has been implemented that allows content to be delivered in markdown for agents while maintaining a full HTML and CSS experience for human readers. This is achieved through content negotiation, where the HTTP Accept header specifies the preferred format, and agents like Claude Code signal a preference for markdown by listing it first in their requests. A middleware checks these preferences and routes markdown-preferred requests to a Next.js route handler, converting rich-text content to markdown while preserving its structure, ensuring that code blocks, headings, and links remain functional. The markdown version significantly reduces payload size, allowing agents operating under token limits to consume more content efficiently. Synchronization between HTML and markdown versions is maintained using Next.js 16 remote cache and shared slugs, ensuring that both versions refresh simultaneously when content updates occur. Additionally, a markdown sitemap helps agents discover available content by listing metadata about each piece, including publication dates, content types, and links to both HTML and markdown versions, enabling agents to select the format best suited for their needs.

Vercel has launched a public bug bounty program on HackerOne for its open-source projects, inviting security researchers to identify vulnerabilities and enhance the security of its widely-used tools, which include frameworks like Next.js and Nuxt, among others. This initiative follows a successful private bug bounty program that began in August 2025, which yielded significant security insights and improvements. By opening the program to the public, Vercel aims to build on its proactive security measures that previously paid out over $1 million to researchers, emphasizing collaboration over confrontation. The program covers all Vercel open-source projects, prioritizing those with the highest potential impact, and provides clear guidelines for participation and reporting. Researchers are encouraged to submit vulnerabilities with detailed reproduction steps via HackerOne, where Vercel's security team is committed to prompt and transparent handling of disclosures to protect the millions of applications relying on these tools.