# Research and open questions

Explore experiments, their limits, and the findings that change how we design agents.

Author: Murali Sid (https://linkedin.com/in/muralisid)
Source: https://www.agenticarchitectureskills.com/research
Published: 2026-08-23
Updated: 2026-10-15
Licence: CC BY-SA 4.0, https://creativecommons.org/licenses/by-sa/4.0/
Cite as: Murali Sid. Research and open questions. Agentic Architecture Skills. https://www.agenticarchitectureskills.com/research

---


Research supports decisions in the guide. This section distinguishes author-reported measurements, external evidence, and investigations that are only proposed.

<TeachingIllustration src="/figures/research/research-cycle.svg" mobileSrc="/figures/research/research-cycle-mobile.svg" title="How research changes the guide" alt="Question: State what is uncertain. Comparison: Use a credible baseline. Experiment: Measure successes and failures. Interpretation: Keep scope and limitations. Guidance: Update affected advice" caption="Planned investigations are labeled as questions, not presented as completed experiments." />

<Cards>
  <Card title="Experiments and results" href="/research/experiments">
    Retrieval, context, and memory findings, including results that did not support the proposed approach.
  </Card>

  <Card title="Questions to investigate" href="/research/questions">
    Specific open questions about representations, investigation loops, and business value.
  </Card>

  <Card title="What changed our advice" href="/research/what-changed">
    Where evidence caused us to narrow or revise a recommendation.
  </Card>

  <Card title="How we evaluate evidence" href="/research/method">
    Baselines, workload scope, source quality, reproducibility, and uncertainty.
  </Card>
</Cards>

## Explore the evidence by subject [#explore-the-evidence-by-subject]

* [Context and retrieval](/patterns): author-reported experiments and practical patterns.
* [Memory measurements](/memory/what-we-have-measured): what was tested and what the results do not establish.
* [Memory benchmarks](/memory/what-the-benchmarks-measure): how comparisons can mislead.
* [Industrial examples from external sources](/research/industrial-examples): documented capabilities and their limits.
* [The fourteen layers](/layers) and their [research findings](/library/layers): detailed findings and dated external sources.

## Where things stand [#where-things-stand]

Each new design example identifies its scope and limitations. A proposed agent extension is not labeled as an operational deployment. Existing experiment results retain the circumstances in which they were obtained; they are not universal performance promises.

The [reference library](/library) contains the deeper explanations. The [architecture](/architecture) and [use cases](/use-cases) show where the evidence affects a decision.