WebML WG Teleconference – 15 January 2026 - 15:00-16:00 UTC

Latest version: https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md


WebML WG Teleconference – 15 January 2026 - 15:00-16:00 UTC
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md#webml-wg-teleconference--15-january-2026---1500-1600-utc>

 Important

⏰ Early morning warning for the West Coast.

See the timezone table ...
San Francisco   Thu, 15 January 2026    07:00
Boston  Thu, 15 January 2026    10:00
London  Thu, 15 January 2026    15:00
Berlin  Thu, 15 January 2026    16:00
Helsinki        Thu, 15 January 2026    17:00
Shanghai        Thu, 15 January 2026    23:00
Tokyo   Fri, 16 January 2026    00:00
UTC     Thu, 15 January 2026    15:00 UTC

Other locations: https://www.timeanddate.com/worldclock/fixedtime.html?iso=20260115T15


🤝 Logistics
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md#-logistics>

  *   Chair: Anssi
  *   Scribe: Anssi
  *   IRC: irc://irc.w3.org:6667/#webmachinelearning

  *   IRC web client: https://irc.w3.org/?channels=#webmachinelearning

  *   Zoom joining instructions: internal-webmachinelearning<https://lists.w3.org/Archives/Member/internal-webmachinelearning/2023Jun/0000.html>
  *   Minutes: https://www.w3.org/2026/01/15-webmachinelearning-minutes.html


🧪 Incubations
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md#-incubations>
🔮 New proposal: Dynamic AI Offloading Protocol
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md#-new-proposal-dynamic-ai-offloading-protocol>

A new proposal to address the challenges related to offloading inference tasks from servers to client devices. Brainstorm use cases and implementation strategies to inform this exploration.

  *   ⨀ webmachinelearning/proposals#15<https://github.com/webmachinelearning/proposals/issues/15>

🧠 Web Neural Network API
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md#-web-neural-network-api>

Tip

Specification<https://www.w3.org/TR/webnn/> | Explainer<https://github.com/webmachinelearning/webnn/blob/main/explainer.md> | Implementation Report<https://wpt.fyi/results/webnn> | Implementation Status<https://webmachinelearning.github.io/webnn-status/>

📍 Candidate Recommendation Snapshot 2026 review
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md#-candidate-recommendation-snapshot-2026-review>

Review the staged snapshot of the upcoming major spec release. Resolve to publish a new CRS.

  *   ⛙ Staged snapshot webmachinelearning/webnn#915<https://github.com/webmachinelearning/webnn/pull/915>
  *   ☰ Release history https://www.w3.org/standards/history/webnn/

⚙️ Implementation experience, from the past to the future
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md#%EF%B8%8F-implementation-experience-from-the-past-to-the-future>

Share implementation experience and learnings from both earlier (webnn-native<https://github.com/webmachinelearning/webnn-native>) and new emerging efforts (rustnn<https://github.com/rustnn/rustnn>, pywebnn<https://pypi.org/project/pywebnn>) coming hot on the heels of the Chromium implementation.

ℹ️ Accelerated context option implementation feedback
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md#%E2%84%B9%EF%B8%8F-accelerated-context-option-implementation-feedback>

Discuss implementation feedback wrt accelerated context option and its interaction with power preferences. Agree on spec changes.

  *   ⨀ webmachinelearning/webnn#911<https://github.com/webmachinelearning/webnn/issues/911>

ℹ️ Floating point accuracy for sin and cos with range constraint
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md#%E2%84%B9%EF%B8%8F-floating-point-accuracy-for-sin-and-cos-with-range-constraint>

Discuss the proposal to define accuracy for sin and cos, any WPT enhancements.

  *   ⨀ webmachinelearning/webnn#914<https://github.com/webmachinelearning/webnn/issues/914>

ℹ️ Drop support of 8-bit integers input for CumulativeSum
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md#%E2%84%B9%EF%B8%8F-drop-support-of-8-bit-integers-input-for-cumulativesum>

Merge PR. Discussed 2025-10-09<https://www.w3.org/2025/10/09-webmachinelearning-minutes.html#fef1>, no concerns recorded for the proposed change.

  *   ⨀ webmachinelearning/webnn#892<https://github.com/webmachinelearning/webnn/issues/892>
  *   ⛙ webmachinelearning/webnn#912<https://github.com/webmachinelearning/webnn/pull/912>

ℹ️ Add minimum data type set and rank range for input, constant, output
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md#%E2%84%B9%EF%B8%8F-add-minimum-data-type-set-and-rank-range-for-input-constant-output>

Review the revised PR and merge unless new blockers. Discuss as needed.

  *   ⨀ webmachinelearning/webnn#896<https://github.com/webmachinelearning/webnn/issues/896>
  *   ⛙ webmachinelearning/webnn#910<https://github.com/webmachinelearning/webnn/pull/910>

ℹ️ Remove conformance tests with negative scale of DQ/Q operators
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md#%E2%84%B9%EF%B8%8F-remove-conformance-tests-with-negative-scale-of-dqq-operators>

Revise the PR per comments and merge. Discuss as needed.

  *   ⨀ webmachinelearning/webnn#879<https://github.com/webmachinelearning/webnn/issues/879>
  *   ⛙ webmachinelearning/webnn#906<https://github.com/webmachinelearning/webnn/pull/906>

ℹ️ Operator chaining
<https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-01-15-wg-agenda.md#%E2%84%B9%EF%B8%8F-operator-chaining>

Revisit operator chaining proposal with new information.

  *   ⨀ webmachinelearning/webnn#106<https://github.com/webmachinelearning/webnn/issues/106>