Breaking news

Top AI Researchers Pivot From Big Tech To Launch Billion-Dollar Startups

Elite Talent Charting a New Course

AI researchers are leaving large technology companies, including Meta and Google, to launch new startups. Founders are using prior research experience to build early-stage companies and attract investor funding. Capital is increasingly directed toward these ventures as investors back teams with established track records in AI development.

Record-Breaking Capital And Strategic Shifts

Former Google DeepMind researcher David Silver has raised $1.1 billion for startup Ineffable Intelligence. Tim Rocktäschel, also previously at DeepMind, is raising close to $1 billion for Recursive Superintelligence. Yann LeCun founded AMI Labs, which reported a $1 billion funding round. The transactions indicate increased funding flows toward AI startups founded by former large-tech researchers.

Strategic Recruitment And Research Diversification

Teams at these startups include former employees of OpenAI, DeepMind, Anthropic, and xAI. Elise Stern, managing director at Eurazeo, said founders with experience in large organizations understand both scaling processes and internal constraints. Focusing in large labs on speed and benchmark performance can limit work in areas such as new architectures, interpretability, and specialized models. Startups are allocating resources to these areas.

Exploiting Niche Opportunities In A Competitive Landscape

Startups, including Ricursive Intelligence and Periodic Labs, are targeting specific use cases. Ricursive Intelligence, founded by former researchers from Anthropic and Google DeepMind, is developing AI tools for chip design. Teams involved in earlier projects, including Google’s AlphaChip, are being reassembled in new companies. These firms are positioning themselves as independent suppliers to semiconductor companies.

Reimagining The Future Of AI

Investment in AI startups reached $18.8 billion in 2026. Funding is concentrated in companies led by researchers with prior experience in large AI labs. Industry discussions are focusing on whether current large language model approaches can support future development. Startups are working on areas including robotics, healthcare applications, and manufacturing systems, with a focus on reliability and model performance.

Conclusion

Movement of researchers from large technology companies to startups is changing how AI projects are funded and developed. The trend is increasing competition across early-stage companies and established firms.

OpenAI Brings Unlimited Text Chats To Free ChatGPT Users

Unlimited Text Conversations Roll Out

OpenAI is removing limits on text-based conversations for all ChatGPT users, following the platform’s recent milestone of surpassing one billion weekly users.

The update introduces GPT-5.6 Luna as the default model for Free and Go users, replacing GPT-5.5.

New “Think” Button For More Complex Questions

Alongside unlimited text chats, Free and Go users will gain access to a new “Think” button, allowing them to use additional reasoning power for more complex queries.

OpenAI noted that separate usage limits will continue to apply to file uploads, image generation, voice features and other multimodal tools.

Faster Responses For Paid Subscribers

The update also brings improvements for ChatGPT Plus and Pro subscribers. They will receive access to GPT-5.6 Sol, an upgraded model designed for everyday tasks such as web research, planning, writing, decision-making and answering questions. According to OpenAI, the model delivers shorter, more robust responses.

The company noted that this version is separate from the GPT-5.6 Sol model used in Codex and Work, which remains unchanged. Plus and Pro subscribers will also receive a new thinking slider, allowing them to adjust how much reasoning the model applies before generating an answer, depending on the complexity of the task.

OpenAI Reports Fewer Errors

According to OpenAI’s internal testing, GPT-5.6 Luna produces 62% fewer factual errors than GPT-5.5 Instant, while GPT-5.6 Sol reduces factual errors by 68%.

The updated GPT-5.6 Sol model is available to Plus and Pro users starting today. The remaining features for Free and Go users will roll out throughout the week, with unlimited text chats and the new Think button becoming available next week.

eCredo
Aretilaw firm
Uol
The Future Forbes Realty Global Properties

Become a Speaker

Become a Speaker

Become a Partner

Subscribe for our weekly newsletter