This site
Artificial Quirks is a working notebook on how agent systems actually behave in production—not demos, not vendor decks.
The industry is shipping recursive loops, multi-model fleets, and “just add RAG” stacks at high speed. The interesting problems are quieter: which knobs actually move cost, when a prettier metric hides a failed job, and how to keep classical software discipline when the runtime is a language model.
What you’ll find here
- Agent runtime design — cell budgets, history working sets, strategy presets instead of model-name
ifs. - Evaluation that measures the job — completion and artefacts alongside cells/turn and token burn.
- Industry notes — architecture trade-offs for GenAI platforms, risk vectors, and operational honesty (ledgers, HITL, exhaustion reasons).
- Math-flavoured systems thinking — when quantitative tools (filters, portfolios, RL) meet messy agent telemetry.
How to read the site
Start with Current series for the recursive language-model loop notes. Use Blog for the full archive. Services is optional advisory; About is credentials—not the product of the site.
Claims stay labeled: observation, inference, provisional. Numbers come with horizons, stop conditions, and limits.
Current series
Recursive language-model loops: three notes on cells, history, and named strategies.
- 1 When Smaller Cells Make Agents Worse
Prompt-only “smaller cells” improved aesthetics and missed completion under a short step budget.
- 2 RLM Is Not Automatically Token-Efficient
History policy plateaus input tokens; cheaper runs still need deliverable checks.
- 3 Strategies, Not Model Ifs
Named YAML strategies beat model-id branches in the agent runtime.
All posts
Topics
-
Agent runtimes
Recursive loops, code cells, tool budgets, and the policies that keep long-horizon agents from burning the ledger.
-
Evaluation & honesty
Measure the deliverable—not only proxy aesthetics. Pair tokens, cells/turn, and step use with FINAL and artefacts.
-
Strategy over model ifs
Named execution strategies, profile bindings, and task overrides—without hardcoding vendor strings in the engine.
-
Platform & risk
Conversational GenAI, RAG, and enterprise constraints: architecture choices that survive audit and operations.
Work with us
Optional advisory: agent runtime reviews, evaluation design, and GenAI platform architecture under real operational constraints.
Recent posts
- Why Build an Agent Harness?
An agent harness turns coding-agent activity into a bounded, inspectable, and repeatable engineering process rather than a stream of opaque terminal outcomes.
- Zero-Spend Concurrency Auditing for Agent Sandboxes
A mock-driven diagnostic method for exercising sandbox resource cleanup without live-model spend.
- Catching Hallucinated Paths in RLM Trajectories
How pre-finalization trajectory hooks catch ungrounded file paths and trigger self-correction turns before final answer submission.
- Grounding Is Not Proof: 100% Code, 0% Plan Quality
Why a 100% task completion rate does not guarantee plan quality, and how pre-finalization checks prevent hallucinated path claims.
- Inside the Sandbox: Unconstrained Shell Execution
Exploring why allowing agents arbitrary host shell access leads to non-deterministic failures and why strict Docker isolation is mandatory for production.
- The Security Cost of Autonomy: Auditing 29 Runs
Empirical security audit results from 29 benchmark study runs evaluating container mount points, path traversal attempts, and write-tool boundaries.
- From REPL Loops to Swarms: Leasing Sandboxed Containers
Designing SwarmLeaseManager to lease container worker nodes dynamically for concurrent file inspection and sub-task execution.
- Targeted Trajectory Repair: Continuations vs Reruns
How provenance-linked continuation child runs inherit parent trajectory history to perform targeted grounding repairs at a 90% reduction in LLM compute costs.
Blog archive