Edition of Wednesday 11 November 2026

Long-horizon agents and durable execution

Comments on “Long-horizon agents and durable execution”

Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked

Kiran Prakash, writing on martinfowler.com, uses a stock-ordering agent to show how durable execution, offered by Temporal, Restate, Inngest, DBOS and Cloudflare Workflows, keeps long-running agents alive across crashes and multi-day waits. The runtime logs each external step and replays the loop after a failure. Retries are at-least-once, so the harness still needs idempotency keys. The wake-up timings come from the authors' own runs on Temporal and Cloudflare, and the article leaves other failure modes, such as code changes during a wait, to harness design.

Read the original at martinfowler.com. The judges saw only its text, with the site, the author and link domains removed; they are shown again after ranking.

Why it ranked

No. 15 of 143 with a score of 0.71. The panel rated it highest on technical depth (0.88) and new information (0.82), lowest on significance (0.63). Placed No. 15, No. 14 by score alone: the fifth AI & machine learning story of the day (−0.040). The placement terms move stories; they are not part of the score.

AI judgeJudge 2claude-haiku-5-5
Substantive, detailed guide to durable execution for long-running agents with code and a worked example; practical for builders, though not a landmark and somewhat long and vendor-adjacent.
AI judgeJudge 3gpt-5.6-luna
A detailed, code-backed engineering treatment of durable execution for long-horizon agents, including replay, retries, idempotency, workflows, memory, prompt injection, and deployment evolution. It is broadly actionable, with modestly speculative conclusions and a minor product mention.

Each judge wrote its line after answering the rubric, with the site and the author still hidden. Every answer behind this is in the table under Score.

Score

0.71total under rubric v2.3

Judges’ mean answer per question, on a scale from 0 to 1. Questions marked minus count against the score.
  • Significance0.63
  • Reader interest0.82
  • Audience breadth0.68
  • Practical relevance0.81
  • Substantiated0.77
  • Technical depth0.88
  • New information0.82
  • Counts against: Promotional0.13
  • Counts against: Overclaim0.20

Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).

QuestionJudge 1Judge 2Judge 3MeanSpreadWeightContribution (weight times mean)
SignificanceHow much does this change what people in tech know or can do?0.710.450.720.630.27+0.24+0.150
Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on?0.780.780.900.820.12+0.26+0.213
Audience breadthHow many people who build technology does it concern?0.750.600.700.680.15+0.13+0.089
Practical relevanceHow useful is it to someone who builds software or tech products?0.860.700.860.810.16+0.22+0.177
SubstantiatedHow well is it backed? Consider whether it is the primary source, and whether its main claims come with data, code, a demo, a reproducible method or citations.0.860.600.860.770.26+0.05+0.039
Technical depthHow technically substantive is it?0.990.750.900.880.24+0.05+0.044
New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)?0.850.700.920.820.22+0.05+0.041
PromotionalIs it mainly marketing with little substance (a press release, ad or sponsored post with no real information)? A launch post that carries real news is not promotional.0.070.250.080.130.18−0.19−0.025
OverclaimDo the headline or conclusions claim more than the body supports?0.230.200.180.200.05−0.11−0.022
Sum of contributions+0.706
Claim-check penalty0.000
Total, as published0.706

Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.

The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.

Ranked #15 on Rubric, 11 Nov 2026
<a href="https://rubric.hrih.si/editions/2026-11-11/15/"><img src="https://rubric.hrih.si/badge/2026-11-11/15.svg" alt="Ranked #15 on Rubric, 11 Nov 2026" width="245" height="28"></a>

Claim check

Claim check: 1 matches source · 7 no source given. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.

  1. No source givenGeneralization

    Temporal, Restate, Inngest, DBOS, and Cloudflare Workflows all implement durable execution.

    The article gives no link or citation for it.

    The article says

    “Temporal, Restate, Inngest, DBOS, and Cloudflare Workflows all implement it”

    The source says

    No source to compare with.

  2. No source givenNumber

    In the authors' runs of Stock Replenisher, waking a workflow took under three seconds and tens of milliseconds of CPU.

    The article gives no link or citation for it.

    The article says

    “In our runs, waking a workflow took under three seconds and tens of milliseconds of CPU”

    The source says

    No source to compare with.

  3. No source givenNumber

    A workflow still in memory woke in 14 ms on Temporal and about 0.1 s on Cloudflare; one rebuilt from the log took about two seconds on Cloudflare and under a tenth of a second on Temporal.

    The article gives no link or citation for it.

    The article says

    “A workflow still in memory woke in milliseconds (14 ms on Temporal, about 0.1 s on Cloudflare); one rebuilt from the log took about two seconds on Cloudflare and under…”

    The source says

    No source to compare with.

  4. No source givenCausal conclusion

    Durable execution runtimes provide at-least-once semantics rather than exactly-once, so a retry after a lost response can place a duplicate order unless an idempotency key is used.

    The article gives no link or citation for it.

    The article says

    “This is known as at-least-once semantics, as opposed to exactly-once. This could result in duplicate orders if not handled properly”

    The source says

    No source to compare with.

  5. No source givenBenchmark result

    When a business rule change (minimum quotes raised from 0 to 4) was deployed while SKU-123 waited, Temporal silently continued with the order under the old rule, while Cloudflare failed and PO 4471 stayed a draft.

    The article gives no link or citation for it.

    The article says

    “On Temporal, SKU-123 silently continued with the order as per the old rule, which the new rule doesn’t permit. On Cloudflare, where our loop checks each recorded decision against the…”

    The source says

    No source to compare with.

  6. Matches sourceGeneralization

    MemGPT and Letta let the harness manage memory in tiers: a small working context the model always sees plus external storage accessed through tool calls.

    The cited source says the same thing.

    The article says

    “MemGPT and Letta let the harness manage memory in tiers: a small working context the model always sees, and external storage like a vector store”

    The source says

    “MemGPT’s OS-inspired multi-level memory architecture delineates between two primary memory types: main context (analogous to main memory/physical memory/RAM) and external context (analogous to disk memory/disk storage”

    Source cited by the article: arxiv.org/abs/2310.08560

  7. No source givenGeneralization

    On Cloudflare, the entity is a Durable Object that reads its current state from storage and has no log to replay.

    The article gives no link or citation for it.

    The article says

    “On Cloudflare, the entity is a Durable Object, which reads its current state from storage and has no log to replay”

    The source says

    No source to compare with.

  8. No source givenGeneralization

    Durable execution does not help with prompt injection threats, and such threats are not unique to agents running for weeks.

    The article gives no link or citation for it.

    The article says

    “None of these threats are unique to agents running for weeks. Durable execution doesn't help with any of them per se”

    The source says

    No source to compare with.

Also covered by

No other coverage was clustered with this story.

Elsewhere

Elsewhere: HN 4 pts, not on front page

Discussion on Hacker News

Recorded for context only, checked 11 Nov 2026, 05:47 UTC. It played no part in the ranking.

Panel for this edition

Four AI agents ranked this edition: three judges scored every story blind to its site and author, and a fourth compared the top stories’ claims with the sources they cite. Summaries are written by AI.

Model versions are pinned for every edition. How ranking works describes rubric v2.3.

Comments

Readers’ comments on this story have a page of their own, ranked by a public rubric rather than by votes.

Comments on “Long-horizon agents and durable execution”

More on Rubric

From the last 7 editions, chosen by shared topic, score and date.

  1. Score 0.74

    Backdooring Codex