Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-08-31 · high
Teams that keep up cite fewer stress spikes at the same error rate, tied to more tools and wider troubleshooting habits.
Why now: Temporal published the findings in its second annual development survey covering April and May.
Agent use in software teams rose 70.8% year over year, according to Temporal's second State of Development report, which surveyed more than 550 engineers, architects and engineering leaders in the US and UK between April 29 and May 25.
The growth number isn't the finding that matters. The report pairs it with a gap: adoption has outrun the infrastructure needed to run agents reliably, concentrated in three places, state, cost, and reliability. That's the part I recognize. Anything that runs for more than a few seconds and has to survive a retry, a crash, or a slow API call needs somewhere to keep its place, and most teams are improvising that with whatever database they already have open.
What separates the teams that aren't drowning in this isn't a better model or a bigger budget. The report ties lower stress at the same error rate to three habits: using more tools, searching more widely when troubleshooting instead of guessing at the first plausible cause, and building internal developer applications rather than relying on ad hoc scripts. In other words, the gap closes with tooling investment, not model swaps.
Temporal runs a durable execution platform, so a report from Temporal concluding that state management is the bottleneck is exactly the conclusion you'd expect it to reach. Take the framing with that in mind. But the underlying complaint, that agents fail in ways ordinary code doesn't and most teams don't have the scaffolding to catch it, matches what I've hit building anything that runs longer than a single request. Worth watching whether next year's version of this survey shows the gap narrowing or just shifting to a new bottleneck once state gets solved.
Each link below shares sources, entities, or timing with this story.
Temporal partners with OpenAI / Shared entity: Teams / Shared topic / Earlier coverage
Linked by a graph relationship (Temporal partners with OpenAI); both cover Teams; overlapping topics (agent, developer).
Temporal partners with OpenAI / Shared entity: April / Shared topic / Earlier coverage
Linked by a graph relationship (Temporal partners with OpenAI); both cover April; overlapping topics (agent, april).
Temporal partners with OpenAI / Shared entity: Vendor / Earlier coverage / Tension
Linked by a graph relationship (Temporal partners with OpenAI); both cover Vendor; earlier Vendor coverage from 2026-08-13.
Temporal partners with OpenAI / Shared entity: April / Earlier coverage
Linked by a graph relationship (Temporal partners with OpenAI); both cover April; earlier April coverage from 2026-04-25.
Copilot uses Teams / Shared entity: Teams / Shared topic / Earlier coverage
Linked by a graph relationship (Copilot uses Teams); both cover Teams; overlapping topics (adoption, agent).
Linked by a graph relationship (Copilot uses Teams); both cover Teams; overlapping topics (cost, developer).
Temporal partners with OpenAI / Shared entity: Vendor / Earlier coverage / Tension
Linked by a graph relationship (Temporal partners with OpenAI); both cover Vendor; earlier Vendor coverage from 2026-07-27.
Temporal partners with OpenAI / Shared topic / Tension
Linked by a graph relationship (Temporal partners with OpenAI); overlapping topics (agent, between, cost); pushes against this story (vs).