|
|
|
|
|
|
When inference outspends payroll, the CFO picks your model - here is the migration playbook.
|
|
JUNE 29TH, 2026 · BY JONATHAN HILDEBRANDT
|
|
Lindy (25 people, 100k+ monthly active users) moved 100 percent of its managed-agent traffic off Anthropic's Claude to DeepSeek v4 Flash on Atlas Cloud on June 24, cutting inference costs roughly 90 percent, per Lindy's migration writeup. CEO Flo Crivello told CNBC inference spend had passed personnel costs and called the switch "a matter of survival." Getting there took offline evals across GLM5.1, Kimi K2.6, and DeepSeek v4 Flash, GEPA-loop prompt re-optimization, and a staged rollout.
Why it matters. One candidate passed offline evals and then failed in production on quality complaints, so the lesson is not "switch models" but "switch with retention monitoring." Claude routes stay live for users who pick them, and Crivello says he switches back the day Anthropic cuts prices.
|
|
Read Lindy's writeup →
|
|
|
|
•
|
Make AI Agents went GA on every paid plan with MCP support. Core at $9/mo for 10k credits per Make's pricing page, roughly 9x lower base cost than Zapier Professional. Try Make →
|
|
•
|
beehiiv MCP v2 write access is now open to every user, not just paid. The only newsletter platform you can drive from Claude, ChatGPT, or Perplexity. Scale at $43/mo switches on the Ad Network and Boosts at 0 percent take rate, per beehiiv's pricing page. Try beehiiv →
|
|
|
|
|
Jonathan Hildebrandt
Co-founder and primary operator of Pondero. Writes the Pondero Brief.
|
|
|
Affiliate disclosure
·
Pondero earns commissions on some links. This does not affect our editorial picks.
|
|
