Evidence

What the numbers mean.

Each claim keeps its source, date, attribution, and limit.

11 public measures · updated July 30, 2026

Numbers

11 claims, each with its limit.

Learning reach

Enrollment and launch measures for three AI courses.

About half

of course enrollments accompanied a W&B account created the same day

An observed same-day association, not a causal conversion claim.

Scope
work
Attribution
direct
Period
2024–2026 · as of 2026-07-30
Source
Internal analytics
SourcesInternal analytics — rounded public summary
1K+

Evals-course signups in roughly 36 hours

Rounded down from the internal launch checkpoint.

Scope
work
Attribution
direct
Period
~36 hours · as of 2025-01-10
Source
Internal analytics
SourcesInternal analytics — rounded public summary

Product adoption

Rounded usage measures for the W&B MCP Server.

200K+

MCP tool calls through July 29, 2026

Rounded down from the July 29 internal analytics checkpoint; direct project usage, not revenue.

Scope
work
Attribution
direct
Period
Through July 29, 2026 · as of 2026-07-29
Source
Internal analytics
SourcesInternal analytics — rounded public summary
More than two-thirds

of weekly active MCP users were returning users

Weekly active-user composition at the cited checkpoint.

Scope
work
Attribution
direct
Period
Week of 2026-06-07 · as of 2026-06-11
Source
Internal analytics
SourcesInternal analytics — rounded public summary
Hundreds

of MCP users re-engaged in a week

Rounded weekly re-engagement checkpoint.

Scope
work
Attribution
direct
Period
Week of 2026-07-20 · as of 2026-07-28
Source
Internal analytics
SourcesInternal analytics — rounded public summary

Program impact

Reach and supported-pipeline measures for public programs.

$4M+

in supported event pipeline across tracked 2024–2025 programs

Supported pipeline across tracked event programs; not revenue, closed ARR, or personal attribution.

Scope
portfolio
Attribution
associated
Period
2024–2025 · as of 2025-10-01
Source
Internal analytics
SourcesInternal analytics — rounded public summary

Archive reach

Counts derived from published project records.

49

dated public event and session references

Unique inventory rows; recordings and mirrors are counted separately.

Scope
collection
Attribution
contextual
Period
2021–2026 · as of 2026-07-30
Source
Public footprint inventory
SourcesPublic footprint counting methodology
19+

hours of public talks and interviews

Derived from 23 unique recording durations; mirrors and excerpts are excluded.

Scope
collection
Attribution
contextual
Period
2021–2026 · as of 2026-07-30
Source
Public footprint inventory
SourcesPublic footprint counting methodology

Method

Three rules.

184 source records support 82 projects and series. Related records remain searchable without being counted twice.

  1. 01

    Count each project once

    Recordings, translations, mirrors, profiles, excerpts, and source pages do not inflate the total.

  2. 02

    Separate contribution from causality

    Individual contribution, program reach, supported pipeline, and organization outcomes remain separate.

  3. 03

    Keep the limit beside the claim

    Each derived measure keeps its period, source, attribution, rounding, and caveat.

Sources

198 sources.

198 sources25—48 shown
  1. 025
    Fine Tuning and Evaluating LLMs for Agentic Use Cases — Part 1Web page · wandb.aiSupports Model adaptation and evaluation for agentic LLMs
    Open source ↗
  2. 026
    Building and Evaluating AgentsVideo · youtube.comSupports Agent development with evaluation in the loop
    Open source ↗
  3. 027
    Understanding the Landscape of the Latest Large ModelsWeb page · machinelearning.co.ilSupports Large model families in 2023
    Open source ↗
  4. 028
    ash0ts — GitHubCode · github.com
    Open source ↗
  5. 029
    When a Plausible Tab Is Wrong: Toraneko Evidence-Level AnalysisWeb page · wandb.aiSupports When a plausible tab is wrong
    Open source ↗
  6. 030
    Tend and Tune Your MARL Experiments with Ray and W&BWeb page · raysummit23.anyscale.comSupports Tracking MARL experiments with Ray and W&B
    Source unavailable
  7. 031
    Mastering Model Customization: Fine-Tuning Azure OpenAI Service Models with W&BVideo · brighttalk.comSupports Fine-tuning Azure OpenAI models with W&B
    Open source ↗
  8. 032
    ash0ts — PyPICode · pypi.org
    Open source ↗
  9. 033
    Mastering Gen AI Applications with W&B — AmericasWeb page · cloudonair.withgoogle.comSupports GenAI application development at Google Cloud Startup School
    Open source ↗
  10. 034
    Emad Mostaque: Stable Diffusion, Stability AI, and What's NextWeb page · wandb.aiSupports Gradient Dissent transcripts and production
    Open source ↗
  11. 035
    groq-webinar-codegen-testingCode · github.comSupports Groq code-generation testing demo
    Open source ↗
  12. 036
    Snowflake Arctic / PubMed RAG — KoreanWeb page · wandb.aiSupports PubMed RAG with Snowflake Arctic and Weave
    Open source ↗
  13. 037
    MLOps World Speaker DirectoryWeb page · mlopsworld.com
    Open source ↗
  14. 038
    Evaluating and Integrating ML Models — MLOps Podcast #213Web page · home.mlops.communitySupports ML model evaluation and integration
    Open source ↗
  15. 039
    Accelerate GenAI Application Development with W&B Weave and NVIDIA — EXS74222Web page · nvidia.comSupports GenAI application development with Weave and NVIDIA
    Open source ↗
  16. 040
    W&B examples Azure medical notes #578Code · github.comSupports Clinical-note generation with fine-tuned GPT models
    Open source ↗
  17. 041
    Accelerating Diffusion with Hugging FaceVideo · youtube.comSupports Faster diffusion workflows with Hugging Face
    Open source ↗
  18. 042
    Weights & Biases — Microsoft Reactor SeriesWeb page · developer.microsoft.com
    Open source ↗
  19. 043
    Architecting and Orchestrating AI AgentsVideo · youtube.comSupports Agent architecture and orchestration series
    Open source ↗
  20. 044
    Mastering MLOps with W&B and Microsoft Phi-2Video · youtube.comSupports MLOps with W&B and Microsoft Phi-2
    Open source ↗
  21. 045
    Prompt Engineering LLMs with LangChain and W&B — JapaneseWeb page · wandb.aiSupports Prompt engineering with LangChain and W&B
    Open source ↗
  22. 046
    wandb-mcp-serverCode · github.comSupports W&B MCP Server for experiments and traces
    Open source ↗
  23. 047
    Evaluating RAG ApplicationsWeb page · luma.comSupports RAG evaluation in practice
    Open source ↗
  24. 048
    W&B Weave arXiv cookbook #2005Code · github.comSupports arXiv paper summarization with Claude and Weave
    Open source ↗