Top AI Product

Five weeks. That’s how long Dreamer existed as a public product before Meta swooped in and acq-hired the entire founding team. On February 18, 2026, the startup formerly known as /dev/agents launched its open beta — a no-code platform where anyone could build, discover, and remix AI agents using plain English. By March 23, Bloomberg reported that Meta had absorbed the team into its Superintelligence Labs division.

The speed is remarkable, but the backstory makes it even more interesting. Dreamer wasn’t some scrappy weekend project. It was backed by $56 million in seed funding at a $500 million valuation, built by a founding team with senior experience at Google, Stripe, Meta, and Xiaomi, and powered by a partnership with Anthropic. The fact that Meta wanted the people more than the technology tells you something about how the AI talent war is playing out in 2026.

What Dreamer Actually Built

Dreamer positioned itself as an “operating system for AI agents” — not just another chatbot wrapper, but a full platform where agents could be created, hosted, and composed together.

At the center of the system sat Sidekick, essentially the platform’s kernel. Sidekick was a personal AI agent that could build other agents on your behalf. You’d describe what you wanted in natural language — say, “I need an agent that monitors my email for flight confirmation and adds the trip to my calendar” — and Sidekick would create a blueprint, check tool access, write the business logic and UI, test it, and deploy it. The whole process typically took 10 to 15 minutes for simple agents.

The platform architecture broke down into three layers:

• Agents (user space): composed of triggers, deterministic code, LLM prompts, sub-agents, interactive UIs, databases, and static files
• Tools (device drivers): pre-built integrations with Google Workspace, GitHub, Linear, speech recognition, image generation, and more
• Sidekick (kernel): the orchestration layer mediating all interactions and managing a persistent memory system with privacy controls

What made Dreamer technically distinctive was its use of Anthropic’s Agent SDK for runtime self-modification. Agents could spawn sub-agents dynamically during execution — creating what co-founder David Singleton described as emergent composability. An agent dealing with news articles could autonomously recruit a read-it-later agent, which would generate audio summaries, which Sidekick would then post to Slack — all without any of that being pre-configured.

The platform also had a gallery for sharing agents, a remix feature for customizing existing agents without building from scratch, and a cross-device experience spanning desktop, iOS, Chrome extension, and email. Android was listed as forthcoming.

The Dream Team (and Their $500M Valuation)

Dreamer’s founding team reads like a greatest-hits album of Silicon Valley engineering leadership:

• David Singleton (CEO): Former CTO at Stripe
• Hugo Barra (CPO): Previously led hardware at Xiaomi and ran Meta’s Oculus VR division. Before that, VP of Android product management at Google
• Ficus Kirkpatrick (CTO): Early Android engineer, former VP of AR/VR at Meta
• Nicholas Jitkoff (Design): Principal designer on Google Chrome, senior roles at Dropbox and Figma

Singleton and Barra first met 18 years ago at Google, building mobile apps before Android even existed. The team of 15 people was backed by a $56 million seed round co-led by Index Ventures and CapitalG (Alphabet’s independent growth fund), with angels including Andrej Karpathy and — here’s where it gets interesting — Alexandr Wang, who would later become Meta’s Chief AI Officer.

That detail matters. Wang invested in /dev/agents as an angel. Then he joined Meta. Then Meta acq-hired the company he had personally backed. The AI world is a very small circle.

Mike Krieger, Instagram co-founder and now head of Anthropic Labs, was another high-profile supporter. “There’s so much intelligence in the models that it deserves not to be trapped in a chat box,” Krieger said when promoting the platform at launch.

From Beta to Meta: Inside the Acq-Hire

On March 23, 2026, Bloomberg broke the news: Meta had hired the Dreamer founding team. The deal was structured as a classic acq-hire — Meta wanted the talent, not the technology. The Dreamer platform itself was not part of the acquisition.

The team is joining Meta Superintelligence Labs (MSL), the division led by Alexandr Wang that houses all of Meta’s AI research and development, including the Llama model family. According to an internal post Wang sent that Monday morning, the Dreamer team will work on AI agents and related projects.

Hugo Barra’s involvement adds another layer. He previously spent years at Meta running the Oculus VR division, so this is effectively a homecoming. The fact that he left Meta, co-founded a startup, raised at a $500 million valuation, launched a product, and then returned to Meta — all within roughly 18 months — captures the current velocity of the AI industry.

What happens to the Dreamer platform and its beta users remains unclear. With the founding team moving to Meta and the technology not included in the deal, the platform’s future is uncertain at best.

Meta’s AI Agent Acquisition Spree

Dreamer isn’t an isolated move. It’s the third agent-related acquisition Meta has made in early 2026, painting a clear picture of where Zuckerberg is placing his bets:

Manus (late 2025/early 2026): Meta spent $2 billion to acquire the autonomous agent platform founded in China — a rare example of a U.S. tech giant buying a Chinese AI startup outright.

Moltbook (March 10, 2026): Meta acquired the viral “social network for AI agents” — a Reddit-style platform where only AI agents could post and interact. Co-founders Matt Schlicht and Ben Parr joined Meta Superintelligence Labs. If you’re curious about what made Moltbook so unusual, we covered it in detail when it first launched and attracted over 150,000 AI agents within days.

Dreamer (March 23, 2026): The acq-hire of Singleton, Barra, Kirkpatrick, and their team, just 13 days after Moltbook.

The pattern is clear: Meta is building an AI agent ecosystem from multiple angles — execution infrastructure (Manus), agent-to-agent social interaction (Moltbook), and consumer-facing agent creation (Dreamer). Combined with Meta’s projected $135 billion in AI capital expenditure for 2026, this is one of the most aggressive agent-focused strategies in the industry.

How Dreamer Compared to Other Agent Builders

Even during its brief public life, Dreamer occupied a distinctive position in the no-code AI agent space. Here’s how it stacked up against the main alternatives:

Dreamer’s biggest differentiator was its OS-level ambition. While platforms like Lindy focus on being a smart assistant and tools like n8n or Make focus on workflow automation, Dreamer aimed to be the runtime environment where agents live, communicate, and evolve autonomously. The Anthropic Agent SDK integration enabling runtime self-modification was something no other consumer-facing platform offered.

The trade-off was maturity. Dreamer was in beta for barely a month. Lindy, n8n, and Make have years of production hardening, thousands of integrations, and established user communities. Whether Dreamer’s architectural vision would have won out over the incumbents’ reliability is a question that will now never be answered — at least not under the Dreamer name.

What This Means for the Agent Platform Market

The Dreamer acq-hire sends a mixed signal to the AI agent startup ecosystem. On one hand, it validates the space — Meta is clearly willing to pay top dollar for agent expertise. On the other hand, it raises the question of whether independent agent platforms can survive when big tech is this aggressive about talent acquisition.

For anyone who was building on Dreamer’s beta, the immediate concern is platform continuity. For the broader market, the takeaway is that the agent OS layer — the infrastructure that hosts, orchestrates, and composes agents — is now a strategic priority for the largest tech companies in the world.

FAQ

What was Dreamer’s pricing model?
Dreamer was in free open beta with a credits-based system. Premium tools from partners were available, and users received free trial credits. The startup also planned a revenue-sharing model for tool creators and was developing agent monetization infrastructure, but the Meta acq-hire happened before any paid tiers launched.

Who were Dreamer’s main competitors?
In the no-code agent builder space, Dreamer competed with Lindy (AI assistant focus), n8n (open-source workflow automation), Make (visual automation), and emerging platforms like MindStudio. However, Dreamer’s “agent OS” positioning made it more architecturally ambitious than most direct competitors. Airtable’s Superagent product also overlapped in the research and multi-agent coordination space.

What happens to Dreamer now that Meta acquired the team?
The acq-hire did not include Dreamer’s technology — Meta acquired the talent only. The future of the Dreamer platform is uncertain. Users who were building agents on the beta should consider migrating to alternative platforms.

Why did Meta want the Dreamer team specifically?
The founding team’s combination of experience — Stripe infrastructure (Singleton), consumer hardware and VR (Barra), Android and AR/VR engineering (Kirkpatrick), and product design at scale (Jitkoff) — maps directly onto the challenges of building consumer-facing AI agent products. Meta’s CEO Mark Zuckerberg has repeatedly emphasized his vision for an “agentic web,” and the Dreamer team has exactly the right background to execute on that vision.

Is Meta building its own agent platform?
All signs point to yes. Between the Manus acquisition ($2B), the Moltbook acq-hire, and now Dreamer, Meta is assembling the pieces for a comprehensive AI agent ecosystem. Combined with the Llama model family and Meta Superintelligence Labs’ resources, Meta appears to be building toward an integrated agent platform — though no official product announcement has been made yet.

You Might Also Like

• Agent Builder by Thesys When ai Agents Stop Talking and Start Showing
• Cloudrouter Gives Your ai Coding Agent its own Cloud Machine and Thats a big Deal
• Google A2ui Agent to User Interface Finally a Standard way for ai Agents to Show you Things
• Stripe Minions 1300 Pull Requests a Week Zero Human Written Code
• Gitnexus Turns Your Codebase Into a Knowledge Graph and Your ai Agent Will Thank you