Braintrust

What drives engagement

Smart avg engagement per post across every post in this window. Gray rows at the bottom are under the sample floor (8+ posts per category, 8+ per format) — too few posts to trust the average.

Best time block: 3pm–6pm UTC. Timing detail is in the Cadence chart above.

Best performing posts

• 2026-08-11 · Company page · Customer Highlight70 eng. (57 likes · 0 comments · 13 shares)

  Eve's legal agents analyze thousands of documents, build medical chronologies, summarize eight-hour depositions, run legal research across c…

  View post →
• 2026-07-30 · Company page · Customer Highlight49 eng. (45 likes · 3 comments · 1 shares)

  Pylon's customers use its support platform to serve their customers, so the stakes for quality are high. Because they have Braintrust in pro…

  View post →
• 2026-08-07 · Company page · Customer Highlight42 eng. (37 likes · 0 comments · 5 shares)

  Cloudflare's dashboard agent spans its entire developer platform, from deploying Workers to debugging production instances. Keeping that muc…

  View post →
• 2026-07-29 · Company page · Product Update38 eng. (31 likes · 4 comments · 3 shares)

  To build long-horizon agents you can trust, you have to supervise the process, not just the final output. Agents that make hundreds of decis…

  View post →
• 2026-08-05 · Company page · Product Update13 eng. (13 likes · 0 comments · 0 shares)

  If you build agents with Cloudflare, you can send traces to Braintrust to see how these agents behave, eval them, and improve them over time…

  View post →

Worst performing posts — settled posts only (published ≥7 days before the data edge)

• 2026-08-04 · Company page · Industry Insight5 eng. (5 likes · 0 comments · 0 shares)

  Eval foundations, from Braintrust. Learn to run evals so you can ship quality agents. - Write deterministic and LLM-as-judge scorers - Compa…

  View post →
• 2026-07-28 · Company page · Industry Insight8 eng. (7 likes · 0 comments · 1 shares)

  Both GLM-5.2 and Opus 4.8 models achieved 83 out of 100 correct answers in our perturbation test, but the cost efficiency gap was substantia…

  View post →
• 2026-08-03 · Company page · Industry Insight9 eng. (8 likes · 1 comments · 0 shares)

  Which design MCP server builds a better front-end page, Figma or Paper? We ran an independent eval in Braintrust with a headless Claude Code…

  View post →
• 2026-07-31 · Company page · Product Update10 eng. (10 likes · 0 comments · 0 shares)

  What's new: - Behavior specs, an open standard for defining and evaling how an agent should behave over an entire trajectory - Pairwise scor…

  View post →
• 2026-08-03 · Company page · Product Update10 eng. (9 likes · 1 comments · 0 shares)

  Early-stage teams building agents know they need evals and observability. But the cost and time commitment often push it down the priority l…

  View post →