All Turtles

All Turtles is a product studio founded by Phil Libin in 2017 that builds and invests in artificial intelligence startups, operating as a distributed team across multiple countries.

All Turtles is a product studio and venture firm founded in 2017 by Phil Libin, previously the CEO of Evernote. The organization focuses on creating and scaling artificial intelligence (AI) companies, combining in-house product development with external startup investment. It operates with a distributed workforce across offices in San Francisco, Tokyo, Paris, and other locations, aiming to address practical problems through AI applications rather than pursuing fundamental research.

The studio's model involves identifying market opportunities, forming teams around them, and providing operational support, design, and engineering resources. All Turtles has launched several ventures, including the AI-powered meeting assistant meetgeek and the mental health chatbot woebot, though the latter later spun out as an independent company. The name "All Turtles" derives from a metaphor about building companies that are "all the way down," emphasizing a hands-on approach to venture creation.

Founding and Leadership

Phil Libin founded All Turtles in March 2017 after stepping down as CEO of Evernote. He was joined by co-founder and chief operating officer jessica-verrilli, who previously worked at Twitter and Google. The initial team included engineers and designers from Evernote, slack, and other technology companies. Libin's vision was to create a studio that could rapidly prototype and launch AI products, avoiding the traditional venture capital model of investing in external founders only.

The company raised an initial funding round of $50 million from investors including emerson-collective and mitsui-fudosan. By 2018, All Turtles had grown to over 100 employees across its global offices. The studio's leadership emphasized a culture of experimentation, with projects evaluated on a quarterly basis for continuation or shutdown.

Business Model

All Turtles operates as a hybrid between a startup incubator and a venture fund. It provides seed funding, typically between $1 million and $5 million per project, in exchange for equity. The studio supplies shared services such as legal, human resources, marketing, and infrastructure, allowing founding teams to focus on product development. Unlike traditional accelerators, All Turtles retains a larger ownership stake, often between 40% and 60%, reflecting its active role in building the company.

The studio also runs an accelerator program for external startups, offering a 12-week curriculum and access to its network. Revenue comes from equity gains upon exits, as well as consulting fees for AI integration projects with established corporations. As of 2020, All Turtles had launched 15 ventures, with two achieving acquisition or significant funding rounds.

Key Products and Ventures

One of All Turtles' earliest products was meetgeek, an AI assistant that automatically transcribes and summarizes meetings. Launched in 2018, it integrated with zoom and microsoft-teams, using Natural language processing to extract action items. Meetgeek gained traction among small businesses, reaching 10,000 users within its first year.

Another notable venture was woebot, a chatbot for mental health support based on cognitive-behavioral-therapy. Initially developed within All Turtles in 2017, it later became an independent company, Woebot Health, raising over $100 million in subsequent funding. Other projects included cove, a customer service automation tool, and hubble, a data analytics platform for e-commerce.

In 2019, All Turtles launched ai-for-good, a non-profit initiative to apply AI to social challenges such as disaster response and education. This program partnered with academic institutions and NGOs, though it remained a small part of the studio's overall portfolio.

AI Approach and Technology

The studio's technical approach emphasizes practical Machine learning applications over cutting-edge research. Teams use existing frameworks such as TensorFlow and PyTorch to build models for Natural language processing, Computer vision, and Predictive analytics. All Turtles does not develop proprietary algorithms or foundational models, instead focusing on product-market fit and user experience.

Projects often leverage Transfer learning and pre-trained models from OpenAI and Google DeepMind to accelerate development. The studio maintains a shared codebase and data infrastructure, allowing teams to reuse components like Data Augmentation pipelines and Model Pruning utilities. This modular approach reduces time-to-market, with typical projects moving from concept to beta in under six months.

Distributed Work Culture

All Turtles was an early adopter of fully remote work, with employees spread across time zones in the United States, Europe, and Asia. The company used slack for communication and asana for project management, holding daily stand-up meetings via video conference. This distributed model allowed the studio to hire talent without relocation costs and to operate 24/7 on development cycles.

The culture emphasized autonomy and accountability, with each venture team having significant decision-making power. All Turtles also hosted an annual "TurtleCon" conference, bringing together employees and external partners for workshops and demos. In 2020, the COVID-19 pandemic validated the remote-first approach, as the studio's infrastructure required minimal adjustment.

Funding and Financials

All Turtles secured its initial $50 million from a group of investors including emerson-collective, mitsui-fudosan, and several unnamed family offices. In 2019, it raised an additional $30 million in a Series A round led by Horizons Ventures. The studio's financial model relies on exits, with its first major success being the sale of meetgeek to cisco in 2021 for an undisclosed amount, reportedly in the tens of millions.

As of 2023, All Turtles had invested in over 20 companies, with a combined valuation exceeding $500 million. However, the studio has not disclosed its internal rate of return or overall profitability. Some ventures, such as cove, were shut down after failing to gain traction, reflecting the high-risk nature of the studio model.

Partnerships and Ecosystem

All Turtles has collaborated with several technology companies to integrate AI into existing products. In 2018, it partnered with Salesforce to develop AI-powered customer relationship tools, though the project was discontinued after a year. The studio also worked with IBM Watson on healthcare applications, and with NVIDIA on GPU-optimized model training.

Academically, All Turtles has sponsored research at MIT CSAIL and Stanford AI Lab, focusing on human-AI interaction. These partnerships provided access to cutting-edge research while giving academics real-world deployment contexts. The studio also participated in OpenAI's early access program for its API, using it to prototype conversational agents.

Controversies and Criticisms

The studio model has faced criticism from some venture capitalists who argue that taking large equity stakes discourages top-tier founders. Additionally, the rapid prototyping approach has been accused of prioritizing speed over quality, leading to products with shallow AI integration. In 2019, a report by techcrunch noted that several All Turtles ventures had pivoted multiple times, raising questions about strategic focus.

Privacy concerns have also arisen, particularly regarding meetgeek's data handling practices. The company faced a complaint in 2020 from the electronic-frontier-foundation about potential violations of gdpr in recording meetings without explicit consent. All Turtles responded by updating its consent mechanisms and limiting data retention periods.

Legacy and Impact

All Turtles has contributed to the growth of the AI startup ecosystem by demonstrating an alternative to traditional venture capital. Its distributed work model influenced other companies, particularly during the pandemic. The studio's focus on practical AI applications helped bridge the gap between academic research and commercial deployment, though it has not produced breakthrough technologies comparable to those from OpenAI or Google DeepMind.

As of 2024, All Turtles continues to operate, with a smaller portfolio than its peak. The studio has shifted toward more selective investments, focusing on AI for enterprise efficiency and climate tech. Its long-term impact remains to be seen, but it has established a template for product studios in the AI space.