PLDI 2026

Basics

Field Value
Dates <2026-06-15 Mon><2026-06-19 Fri>
Venue Limelight Boulder, Boulder, Colorado
Host ACM SIGPLAN
URL pldi26.sigplan.org
Streams PLDI 2026 Live Streams (ACM SIGPLAN)
Format Multi-track with co-located events, workshops, tutorials
Framing PL research, compilers, verification; PAgE workshop on agentic engineering

Co-located conferences

  • ISMM 2026 – International Symposium on Memory Management
  • LCTES 2026 – Languages, Compilers, and Tools for Embedded Systems

Workshops

Workshop Full name Notes
ARRAY 2026 Array-oriented Programming  
CP 2026 Choreographic Programming Kuper keynote; correct-by-construction distributed
EGRAPHS Equality Graphs E-graph rewriting, optimization
PAgE 2026 Principles of Agentic Engineering Formal foundations for agent safety/reliability
PLMW PL Mentoring Workshop  
PROPL Programming for the Planet  
SOAP 2026 State of the Art in Program Analysis  
-- Tom Ball at Sixty Celebration session

Tutorials

  • Writing Performance-Portable Kernels Simplified with Helion
  • A guided tour through Oxidized OCaml
  • Implementing an Integrated Development Environment
  • Mechanized Specifications for Real-World Programming Languages
  • ACT: End-to-End Compiler Infrastructure for Emerging AI Accelerators
  • Deep dive into the AWS Nitro Isolation Engine

PAgE 2026: Principles of Agentic Engineering

Full page: program, accepted papers, recording

The workshop most directly connected to current research. 10 accepted papers, 2 keynotes, chaired by Shraddha Barke. Mon 15 Jun, Meadows B. Recording.

Research papers

Wed 17 Jun research papers relevant to our threads (type systems, contracts/ verification, distributed/choreographies, abstract interpretation + LLMs, dependent types) — with DOIs/preprints in research-papers/references.bib and the speaker cross-links (Singh→PAgE, Sorensen/Martínez→PAgE kernels, Das→CP, Pierce/Jhala/Rompf→four-ways).

Why

PLDI sits at the intersection of formal methods and practical compiler engineering. Three reasons this year matters:

  1. PAgE workshop. First SIGPLAN-affiliated venue explicitly framed as "agentic engineering." The formal-methods community claiming this term has downstream consequences for how agent governance gets specified. The elenctic-spec framing and Seven Concerns layering should be testable against whatever vocabulary PAgE settles on.
  2. Nitro Isolation Engine tutorial. AWS Nitro is the production hypervisor-isolation analog to Antithesis. The tutorial is a data point on how far hardware-backed isolation has moved since BugBash.
  3. Egraphs workshop. E-graph rewriting is the optimization substrate beneath equality saturation. Relevant to any spec-refinement system that needs to normalize equivalent representations – the same problem elenctic-spec faces at the methodology layer.

Refutation

The PAgE framing for this event is wrong if:

  • The workshop is dominated by LLM-agent benchmarking rather than formal foundations (safety/reliability becomes eval leaderboards).
  • The "agentic engineering" vocabulary diverges completely from the Seven Concerns framing, with no productive mapping between them.
  • The Nitro tutorial is marketing rather than technical substrate.

In those cases PLDI reduces to a compiler conference with an adjacent agent workshop that didn't converge on useful primitives.

Concern mapping

Event/Talk Concern Claim under test
PAgE workshop (overall) L3, L7 PL community can formalize agent governance primitives
Grossman/Song/Zorn committee L3 Formal methods vocabulary for agents will be PL-native
Nitro Isolation Engine L1, L5 Hardware isolation composes with agent sandboxing
Egraphs workshop L2, L3 Equality saturation applies to spec normalization
Mechanized Specifications L3 Mechanized specs are the reader-as-spec-emitter generalized
Oxidized OCaml tutorial L1, L2 Rust ownership model transfers to ML family

Related