29 Early-Stage Companies: Who Will Be the Next Unicorn? Top Silicon Valley Entrepreneurs Reveal Their List.
29 early-stage companies, who will be the next unicorn? Top Silicon Valley entrepreneurs provide a list
Content Summary
The next industry-changing company may not have entered most people’s view yet. Recently, Silicon Valley tech media Newcomer together with venture capital firm Felicis released a report called “Emergent”.
The next industry-changing company may not have entered most people’s view yet.
Recently, Silicon Valley tech media Newcomer, together with venture capital firm Felicis, released a startup list called “Emergent,” inviting many well‑known founders to recommend early‑stage companies they favor.
Vercel founder Guillermo Rauch, Runway founder Cristóbal Valenzuela, ElevenLabs co‑founder Mati Staniszewski, and LinkedIn co‑founder Reid Hoffman, among others, participated in the recommendations, and ultimately 29 companies were selected.
All of these companies are at a financing stage no later than seed round.From agent payments, security, and management to robotics, AI healthcare, biotech, and music/film production, the business scope of the 29 companies is quite diverse.At a time when AI startups are proliferating, this list offers a window into the entrepreneurial trends in Silicon Valley.
(Source: Newcomer)
Agent Infrastructure: AI Begins to Require Payment, Security, and Management Systems
Among 29 companies, tools and infrastructure around agents account for a substantial proportion. As agents start shopping on behalf of users, operating enterprise software, and even executing complex tasks independently, needs for payment, security, permission management, and reliability are gradually emerging.
Payments are one of those new markets. Agentcard, founded by Karen Serfaty and Felipe Abello, aims to enable agents to complete online transactions autonomously, including purchasing goods and subscribing to services. Vercel founder Guillermo Rauch is bullish on this direction, because as AI moves from offering suggestions to actually executing tasks, payment capability becomes a crucial link between digital services and commercial activity.
As agents gain more operational permissions, security management also becomes more complex. Startup Arrakis Security can detect agents running inside enterprises, analyze their behavior, and block or revoke actions when risks are identified. ElevenLabs founder Mati Staniszewski notes in his recommendation that traditional security tools were primarily designed around human users; now that AI is beginning to autonomously perform tasks in customer management and financial systems, security frameworks need to adapt accordingly.
LinkedIn co‑founder Reid Hoffman focuses on operation records and trust issues.The startup Mnemom he favors can generate verifiable records for each action of an Agent, compare actual behavior with corporate instructions, and enable managers to trace the execution process of tasks.
(Source: mnemom)
Similar needs also appear in the reliability of agents. The startup Convergent tries to find failure cases from actual runtime logs, turn them into evaluation environments, and then use programming agents to fix the issues. It has already launched services in legal, finance, human resources and other fields, focusing on solving the problem of agents repeatedly making mistakes in real-world scenarios.
Another company, Ampersand, goes further inside enterprises by analyzing employee actions and business processes to build a continuously updated operational model for planning automation tasks, evaluating agents, and improving workflows. Its founding team comes from Palantir, OpenAI, and Google. Crosby founder Ryan Daniels also recommends both companies, valuing their understanding of real enterprise workflows.
Beyond security and reliability, model selection and inference cost are also forming an independent market.The startup Martian has developed a model routing system that selects the appropriate large model in real time based on task requirements, helping enterprises control inference expenses. Another company, Conway, has built a predictive foundation model that uses enterprises’ real‑time business data for scenarios such as product recommendation, pricing, and credit assessment.
A group of companies are also enhancing agents’ abilities to acquire information, use software, and complete development tasks.
For example, the startup Armature records agents’ interactions with products via the MCP service, Claude connector, and ChatGPT app, helping developers understand how users engage with software; Attention Engineering collects contextual data generated when users operate computers, providing richer information for personalized AI; Tako organizes data from various sources into a format that language models can understand and query, assisting users in performing complex analyses.
Peak focuses on the new bottlenecks brought by AI programming. As code generation speeds increase, continuous integration and automated testing need to keep up. It provides a faster code build environment, and references claim it has run millions of continuous integration tasks.
(Source: DeepTech illustration)
According to Felicis founder Aydin Senkut, the future Agent economy will require extensive supporting infrastructure.In an interview with Newcomer, he noted that as millions—or even billions—of Agents begin to search for information, use services, and execute tasks via APIs, the importance of identity management, security, and data infrastructure will continue to rise.
From payment and security to model routing and runtime evaluation, these companies are building a new supporting system around the practical use of agents. As enterprise deployment scales up, these niche demands may also give rise to new infrastructure firms.
AI-native applications: moving from generic chat tools to specific work scenarios
Beyond infrastructure, another wave of entrepreneurs is bringing AI into everyday work for businesses and individuals.Compared with generic assistants, they target more specific use cases, and the product value is easier to measure through saved time and completed tasks.
Corporate event planning is one example. Nowadays, founded by Anna Sun and Amy Yan, has developed an Agent that can help companies organize events, handling everything from venue search and vendor outreach to price comparison and negotiation of terms, all delegated to AI. According to ElevenLabs founder Mati Staniszewski, companies such as Google, Amazon, and Notion are already using the service.
This type of work is characterized by cumbersome processes, requiring extensive email communication and multi‑party coordination, yet the final decisions are still made by humans.Nowadays aims to automate highly repetitive tasks so that event managers can focus their energy on solution selection and on‑site execution.
Another company, Hercules, applies AI to enterprise software development. Users can create business applications such as CRM systems or inventory management systems simply by conversing with an Agent. It uses an open‑weight model, aiming to provide software generation capabilities at a lower cost. Fireworks AI founder Lin Qiao is optimistic about this model, believing it can lower the barrier for companies to develop custom software.
Products aimed at individual users are also exploring new markets. Tomo, founded by Justin Quan and Raymond Chen, is developing an AI assistant for handling daily tasks. Josh Tobin, who recommends it, specifically noted that the product is gaining users beyond the Silicon Valley tech circle, showing that AI applications for ordinary consumers are gradually expanding their audience.
(Source: Tomo AI)
However, personal AI assistants are becoming one of the most fiercely contested tracks.In recent months, tech giants such as OpenAI, Meta, Apple, and Google have been pushing personal agent products, competing for the AI entry point in users’ daily lives—from managing schedules and handling emails to shopping and cross‑app operations.
For startups like Tomo, the upcoming competition may become even tougher. Large tech companies already have massive user bases and entry points such as operating systems, social platforms, or office software, making it easier to integrate personal agents into existing products.
The other two companies are turning their attention to office and organizational management. Scuba, co‑founded by former Slack CPO Noah Desai Weiss, is exploring AI‑native office products. Abridge founder Shiv Rao values his team’s experience in collaboration software. SkillBench focuses on employee skill development in the AI era, helping companies identify tasks suitable for automation while cultivating employees’ professional expertise and judgment, and has been endorsed by Reid Hoffman.
(Source: DeepTech illustration)
Hardware and Robotics: AI Moves from Screens to the Physical World
Compared with the software sector, robot startups have always faced higher R&D costs and more complex commercialization challenges.But as models make progress in perception, learning, and task execution, investors are beginning to reassess this market.
Some entrepreneurs choose to start with robot software. Enigma, founded by Jonathan Jacobi and Gal Niv, aims to develop AI systems that work across different robot platforms, giving robots stronger intelligence while lowering operational difficulty. Cartesia founder Karan Goel is optimistic about its universal software approach, because different types of robots have long required hardware‑specific development, and generic software could reduce redundant work.
Icarus Robotics targets physical labor in space. It was founded by Ethan Barajas and Jamie Palmer, and is trying to develop AI robots capable of performing tasks in the space environment. Tines founder Eoin Hinchy believes that labor costs in space operations are extremely high and are tightly constrained by environmental and safety conditions, creating a clear need for automation.
(Source: Icarus Robotics)
Traditional industries also hold opportunities.Agrippa Industries, a startup recommended by Snorkel AI founder Alex Ratner, is exploring the U.S. maritime logistics sector. Although the original list disclosed limited product details, maritime shipping involves complex transport scheduling and operations management, offering substantial space for automation technologies.
Compared with industrial robots, the development path of personal AI hardware is different.Ghost is developing a personal AI computer called Core.It is a screenless device that can connect to a user’s apps, files, and smart home system, running AI models locally to continuously handle personal tasks.
Mirai Labs, which also focuses on local computing, offers an on‑device inference engine and development toolkit, allowing text, speech, and vision models to run directly on phones or laptops. It was founded by entrepreneurs who previously helped create Reface and Prisma, and was recommended by ElevenLabs founder Mati Staniszewski.
Running on the device helps reduce cloud inference costs and network latency, and provides more options for handling personal privacy data. As open‑weight models become more capable, consumer‑grade hardware is becoming a key platform for AI applications.
Two companies have taken a more specialized route. Alljoined applies machine learning to non‑invasive brain‑computer interfaces, attempting to decode what a person sees from EEG signals. Runway founder Cristóbal Valenzuela is optimistic about improving BCI capabilities by scaling data and models.
Mistos applies AI to power infrastructure engineering design, aiming to shorten the design and delivery cycle required for new power facilities. As AI data center construction continues to expand, the importance of power supply and related engineering services is also increasing.
(Source: DeepTech illustration)
Felicis founder Aydin Senkut recalled in an interview that the firm has been involved in robot investments for nearly 15 years, but early projects repeatedly failed. In the past, robots mainly relied on pre‑written programs to perform actions, making it difficult for them to learn quickly through observation and experience like humans. Recent advances in AI technology have begun to change this situation.
He is especially focused on the supply chain behind robots.Beyond the robots themselves, training data, inference systems, actuators, component manufacturing, energy supply, and deployment services could all form new markets as robots scale up.In his view, the truly important validation comes from industrial sites—companies willing to deploy robots in complex production environments is the key signal that the technology is moving toward commercialization.
AI in Healthcare and Biotechnology: From Clinical Decision‑Making to Real‑World Experiments
Healthcare and biotech are another area that Felicis is especially bullish on. The Emergent list includes three companies directly related to this, each approaching it from clinical assistance, biological experimentation, and specialized model development.
In the clinical space, Almanac Health aims to improve the accuracy and trustworthiness of AI medical tools. Founded by Cyril Zakka, the company is dedicated to grounding clinical AI responses in reliable medical evidence, helping physicians access and use professional information more effectively. Glean founder Arvind Jain is optimistic about its approach of combining model capabilities with medical knowledge.
In the realm of biological experiments, startup Instance Bio is working to increase the efficiency of large‑scale molecular testing.Founded by David Scott and Shawn Morel, it conducts parallel experiments to analyze the interactions of a massive number of protein‑binding molecules with various targets.
According to the referral from Adept founder David Luan, researchers can submit candidate protein sequences and targets via an API, after which Instance Bio performs interaction analysis in a single large‑scale parallel reaction. This approach is expected to reduce the time and resources required to test numerous molecular combinations one by one, and also provide biological models with more real experimental data.
The third company, Phylo, founded by Kexin Huang, Yuanhao Qu, Jure Leskovec, and Le Cong, is dedicated to customizing open‑weight models for biological research. Fireworks AI founder Lin Qiao believes this approach can help researchers obtain models suited to specific biological tasks at lower cost, while retaining greater autonomy and flexibility for adjustments.
(Source: DeepTech illustration)
In interviews with Newcomer, Aydin Senkut repeatedly emphasized the importance of experimental validation. He believes that AI can already generate a large number of scientific hypotheses, but the capacity for real‑world experiments and testing remains limited.Increasing experimental throughput and shortening validation cycles will become crucial conditions for AI to further advance biological science.
The capital market is also increasing its investment. Felicis, citing Crunchbase data, says that as of October 30, 2026, global AI healthcare and biotech financing this year reached $20.2 billion, already surpassing the full‑year $15.8 billion in 2025.
AI Creative Tools: From Generating Assets to Full Production Workflows
The last three companies focus on animation, music, and film. Compared with early products that emphasized one‑click generation of images, videos, or audio, they pay more attention to the actual workflows of professional creators.
Runway founder Cristóbal Valenzuela also recommended Cartwheel and Resin. The former, founded by Jonathan Jarvis and Andrew Carr, can convert video, text descriptions, and existing motion assets into 3D character animation and integrate with existing animation pipelines, helping creators reduce repetitive work.
Resin, on the other hand, moves into music production. Founded by Pedro Tsividis, it can generate new instrumental parts—such as bass or strings—based on the track a musician is working on, allowing creators to quickly audition different arrangement ideas. Valenzuela believes tools like this help creators express ideas more easily and share a similar product philosophy with Runway.
In contrast, Wonder Studios’ business model is closer to that of a film production company. This AI‑native studio based in London, founded by Xavier Collins and Justin Hackney, creates commercial content for brands, record labels, and film organizations while also developing original works.
ElevenLabs founder Mati Staniszewski mentioned in his recommendation that Wonder Studios has been involved in projects such as singer Lewis Capaldi’s music video, and has collaborations with organizations like DeepMind and Universal.The company aims to combine AI production technology with the professional talent and production experience of the traditional film industry, reducing production costs while gradually building original content and intellectual property.
（Source: DeepTech illustration）
Looking back at this list, the 29 companies have diverse technology focuses and business models, but several startup trends are becoming clear: Agent infrastructure is forming a new niche market, robots and edge devices are attracting capital, and the medical, biotech, and creative industries are seeing more specialized AI products.
Aydin Senkut noted in an interview that to find the next generation of great companies, one must spot opportunities that have not yet become industry consensus. He is especially focused on areas that could generate massive demand in the future, thereby supporting long‑term growth.
These companies are currentlystill in the early stage, and there is a long way to go before they truly build sustainable businesses. However, this list already shows how a group of successful Silicon Valley entrepreneurs assess the next wave of technology opportunities. Who will become the next unicorn,the market will ultimately decide.
Get MirrorSpeed
Get MirrorSpeed