Imagine AI
Dashboard

Kyle Corbitt

Head of OpenPipe Team · OpenPipe · 11.0K followers · LinkedIn benchmark, all time (original)

OpenPipe company report →
data ends 2026-03-10

Report card — all time

Avg eng/post

92

Median eng/post

36

Posts/wk

0.5

Total eng.

5.3K

Trend (monthly — bars are posts, line is avg engagement per post)

Top categories

  • Product Update — smart avg 74.1 across 23 posts (2.1× typical)
  • Event — smart avg 74 across 8 posts (2.1× typical)
  • Industry Insight — smart avg 67.1 across 18 posts (1.9× typical)

Top posts

  • 2025-07-11 · Kyle Corbitt · Product Update2.0K eng. (1.8K likes · 128 comments · 143 shares)

    We just cracked the code on making AI agents actually reliable. For years, only frontier labs like OpenAI and xAI could use reinforcement le

    View post →
  • 2025-09-03 · Kyle Corbitt · Celebratory Announcement650 eng. (547 likes · 93 comments · 10 shares)

    I have a huge announcement to make: OpenPipe is joining CoreWeave! This has been a huge journey, and it is far from over. We’ll continue ser

    View post →
  • 2024-08-19 · Kyle Corbitt · Industry Insight182 eng. (163 likes · 16 comments · 3 shares)

    So we’ve got ~6 labs with models at the GPT-4 level, and nobody has meaningfully surpassed it. OpenAI had to scrap/restart their GPT-5 train

    View post →
  • 2025-08-06 · Kyle Corbitt · Product Update167 eng. (145 likes · 12 comments · 10 shares)

    Announcing MCP•RL: teach your model how to use any MCP server automatically using reinforcement learning! Just connect any MCP server, and y

    View post →
  • 2025-10-08 · Kyle Corbitt · Product Update136 eng. (125 likes · 7 comments · 4 shares)

    🚀 Announcing Serverless RL — train agents faster and cheaper with zero infra. On our ART‑E benchmark (Qwen 2.5 14B), we saw ~40% lower trai

    View post →
2024-08-192026-08-27 · Benchmark by Imagine AI · methodology at benchmark.imagineai.me/dashboard/methodology