The inaugural ACM Conference on AI and Agentic Systems (CAIS 2026) concluded its four-day run in San Jose, California this week, cementing itself as the new premier venue for rigorous research on compound AI architectures and agentic systems. The conference drew researchers from over 115 academic and industry institutions across three days of peer-reviewed papers, live system demonstrations, and high-profile keynotes โ€” with registration hitting capacity well before opening day, leaving would-be attendees relegated to a waitlist.

A Keynote Roster Worth the Trip Alone

The speaker lineup was stacked. Andy Konwinski, co-founder of Databricks and Perplexity AI and founder of the Laude Institute, took the stage to discuss open-source AI research funding through programs like Moonshots and Slingshots โ€” including the Terminal-Bench benchmark his organization backs, which has become an industry standard for measuring agent command-line performance. Thariq Shihipar, a core builder of Claude Code at Anthropic, delivered technical insights on prompt caching, tool design, and what he calls "unhobbling" โ€” concepts that have quietly shaped how thousands of developers interact with LLMs in practice. Stanford professor Percy Liang rounded out the headliners, bringing his work on the Marin project (frontier models built fully in the open), the HELM benchmarking framework, and coining the term "foundation models" itself.

The Numbers Behind CAIS 2026

The conference ran May 26โ€“29 with a workshops-and-tutorials day bookending three days of main programming. Researchers presented 63 peer-reviewed papers alongside 46 live system demos โ€” working implementations shown by their creators rather than slides. Participating institutions spanned heavy hitters like Google, Meta, OpenAI, Microsoft, NVIDIA, and MIT, alongside universities including Stanford, UC Berkeley, Carnegie Mellon, Princeton, and EPFL. The scope signals that agentic AI research is no longer a niche subfield โ€” it's a full ecosystem attracting the entire stack of players building the future.

Partnership With AI Engineer World's Fair

CAIS 2026 also announced a cross-pollination deal with the AI Engineer World's Fair (June 29โ€“July 2 at Moscone West in San Francisco). Papers earning an Industry Spotlight or Operational Experience designation will be invited to present at both venues โ€” giving authors ACM peer-review credibility plus a stage in front of 6,000+ practicing engineers. It's a smart move that bridges the gap between academic rigor and production reality, which is exactly where agentic systems research needs to go next.

Hotel Details and Logistics

For those who missed it, CAIS was headquartered at the DoubleTree by Hilton San Jose (2050 Gateway Place), with a group rate of $159/night plus tax. The hotel offered complimentary airport shuttle service from SJC and discounted parking at $10/day โ€” practical details that conference organizers clearly put thought into. The room block deadline was extended to May 15, suggesting last-minute demand outpaced initial estimates.

What This Means for the Field

ACM CAIS is positioning itself as a high-signal forum for compound AI systems research โ€” architectures built from multiple components with smart inference-time scaling, verifier-based designs, RAG pipelines, and multi-agent frameworks. The emphasis on reproducibility and systematic verification reflects a maturing field that no longer wants to just ship demos; it wants deployable, reliable systems. With this kind of institutional backing and industry buy-in from day one, CAIS looks like it's here to stay.

Key Takeaways

  • Registration hit capacity early โ€” agentic AI research has serious mainstream traction now
  • Keynote speakers (Konwinski, Shihipar, Liang) represent the full spectrum from open-source infrastructure to production developer tooling to academic benchmarking
  • 63 papers and 46 demos across 115+ institutions signals broad ecosystem participation
  • The partnership with AI Engineer World's Fair bridges academic rigor and engineering practice

The Bottom Line

ACM CAIS 2026 wasn't just another academic conference โ€” it was a statement that agentic systems research has graduated from proof-of-concept territory to a legitimate, rigorous discipline attracting top talent across industry and academia. If you missed your window, start circling CAIS 2027 now.