Making agents fast – what I've learnt over the last 12 months
Comments on “Making agents fast – what I've learnt over the last 12 months”
Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked
Martin Alderson, drawing on a year of optimising agentic workflows, argues that inference speed is only part of agent latency. By his own rough illustration, halving inference time cuts about 20% off a task when tools, provisioning and suspend/resume stay unchanged. He ranks thinking budgets as the first thing to check, then tool-call traces, slow shell startup, polling loops, disk and CPU choice, and microVM resume. The figures are illustrative and unsourced, but the checklist is practical for anyone running agents in sandboxes.
Why it ranked
No. 21 of 59 with a score of 0.64. The panel rated it highest on practical relevance (0.84) and new information (0.78), lowest on substantiated (0.38). Placed No. 21, No. 17 by score alone: the sixth AI & machine learning story of the day (−0.060). The placement terms move stories; they are not part of the score.
- AI judgeJudge 2claude-haiku-5-5
- Practitioner lessons on agent latency: TTFT vs tok/s, thinking budgets, tool-call and sandbox overhead. Useful and detailed but anecdotal, unquantified and not a primary study, so moderate significance and substantiation.
- AI judgeJudge 3gpt-5.6-luna
- A technically grounded practitioner account of optimizing agent workflows across inference, tools, hardware, sandboxes, and suspend/resume. It offers actionable heuristics, but limited quantitative evidence and overlaps somewhat with recent agent-performance coverage.
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
0.64total under rubric v2.3
- Significance0.48
- Reader interest0.73
- Audience breadth0.62
- Practical relevance0.84
- Substantiated0.38
- Technical depth0.75
- New information0.78
- Counts against: Promotional0.04
- Counts against: Overclaim0.17
Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).
| Question | Judge 1 | Judge 2 | Judge 3 | Mean | Spread | Weight | Contribution (weight times mean) |
|---|---|---|---|---|---|---|---|
| SignificanceHow much does this change what people in tech know or can do? | 0.55 | 0.35 | 0.55 | 0.48 | 0.20 | +0.24 | +0.116 |
| Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on? | 0.67 | 0.70 | 0.82 | 0.73 | 0.15 | +0.26 | +0.190 |
| Audience breadthHow many people who build technology does it concern? | 0.65 | 0.50 | 0.72 | 0.62 | 0.22 | +0.13 | +0.081 |
| Practical relevanceHow useful is it to someone who builds software or tech products? | 0.93 | 0.80 | 0.78 | 0.84 | 0.15 | +0.22 | +0.184 |
| 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. | 0.26 | 0.40 | 0.48 | 0.38 | 0.22 | +0.05 | +0.019 |
| Technical depthHow technically substantive is it? | 0.86 | 0.70 | 0.68 | 0.75 | 0.18 | +0.05 | +0.037 |
| New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)? | 0.84 | 0.80 | 0.70 | 0.78 | 0.14 | +0.05 | +0.039 |
| 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. | 0.06 | 0.03 | 0.02 | 0.04 | 0.04 | −0.19 | −0.007 |
| OverclaimDo the headline or conclusions claim more than the body supports? | 0.18 | 0.20 | 0.12 | 0.17 | 0.08 | −0.11 | −0.018 |
| Sum of contributions | +0.641 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.641 | ||||||
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.
<a href="https://rubric.hrih.si/editions/2026-11-01/21/"><img src="https://rubric.hrih.si/badge/2026-11-01/21.svg" alt="Ranked #21 on Rubric, 1 Nov 2026" width="238" height="28"></a>Claim check
Claim check: 8 no source given. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.
No source givenNumber
Halving inference time only takes 20% off the total task time when provisioning, tools and suspend/resume stay the same (the author's rough numbers).
The article gives no link or citation for it.
The article says “halving inference time only takes 20% off the task when provisioning, tools and suspend/resume stay the same”
The source says No source to compare with.
No source givenCausal conclusion
A 'fast' tier model with thinking set to maximum or xhigh will be vastly slower than a 'slow' tier model with low or medium thinking.
The article gives no link or citation for it.
The article says “A "fast" tier model with thinking set to maximum or xhigh will be vastly slower than a "slow" tier model with low or medium thinking”
The source says No source to compare with.
No source givenVersion
OpenAI offers a 'flex' tier alongside its standard and 'fast' tiers, letting users trade speed for cost.
The article gives no link or citation for it.
The article says “OpenAI offers a 'flex' tier alongside their standard and 'fast' tiers. These allow you to trade speed for cost”
The source says No source to compare with.
No source givenGeneralization
Between providers of the same model there is an enormous difference in speed.
The article gives no link or citation for it.
The article says “between providers of the same model1, there is an enormous delta in speed”
The source says No source to compare with.
No source givenCausal conclusion
Agents running xargs -I{} over large numbers of files cause hundreds of thousands of new process invocations, which is painfully slow especially in sandboxed environments.
The article gives no link or citation for it.
The article says “an agent running xargs -I{} over large amounts of files, which causes hundreds of thousands of new process invocations”
The source says No source to compare with.
No source givenNumber
Local NVMe disks may offer many orders of magnitude faster IO performance than networked disks.
The article gives no link or citation for it.
The article says “Having faster (local) NVMe disks may offer many, many orders of magnitude faster IO performance”
The source says No source to compare with.
No source givenGeneralization
Zen 4c/5c CPU cores can be much slower than the full-sized cores.
The article gives no link or citation for it.
The article says “the Zen 4c/5c cores can be much slower than the full sized ones”
The source says No source to compare with.
No source givenGeneralization
Checking thinking budgets is overwhelmingly the lowest-hanging fruit for speeding up agents.
The article gives no link or citation for it.
The article says “Check thinking budgets! This is overwhelmingly the lowest hanging fruit”
The source says No source to compare with.
Also covered by
No other coverage was clustered with this story.
Elsewhere
Elsewhere: HN 2 pts, not on front page
Recorded for context only, checked 1 Nov 2026, 05:47 UTC. It played no part in the ranking.