Evidence (272 claims)

In coding tasks, low agreeableness leads to large communication shifts that have little effect on milestone completion.

Participants' IAT scores were predictive of the time they spent in human-AI collaboration.

Frontier proprietary models achieve near-zero success under GUI-based interaction, whereas COM-based execution yields substantial immediate gains.

The same observation is seen with the amount of changes (e.g., code churn, number of modified files) and with the efforts to merge an agentic PR (e.g., merge time and number of comments).

Claude Code completed the pipeline in ~3.4 minutes with silent deviations from the specification, while Codex required ~16 minutes across explicit self-correcting restarts, including an unsolicited performance optimization of the matched filter inner loop.

AI assistance can generate a deceptive productivity signature: average completion times fall because AI tools typically supply a fast first draft, yet workflow-level performance can deteriorate when a subset of AI errors escapes review and returns as costly downstream rework.

Across 78 endpoints, the same model on different endpoints differs in tail latency by an order of magnitude.

Provisioned Throughput delivers the lowest latency at low concurrency but saturates its reserved capacity above approximately 20 concurrent users.

Wall-clock time can be reduced to O(√E) through team parallelism, but total human effort remains O(E).

For knowledge-intensive workers whose intellectual capital spans tens of thousands of files, the CAD constitutes a qualitative threshold in AI usefulness: below it, the cognitive burden of context curation falls on the human, reproducing the inefficiencies AI is meant to eliminate.

Human interfaces define throughput limits in areas such as prompt engineering, data-stream curation, adjudication of model outputs, and the orchestration of hybrid automation workflows including robotics, scraping, and digitization.

Current machine learning models commonly require large and well-annotated datasets, and the annotation process often becomes a bottleneck with increased complexity leading to higher chances of human errors.

Reconstructing this context can consume an estimated 5,000-20,000 tokens per session.

Extensive experiments across frontier models and agentic systems reveal that even the best-performing configuration (Mini-SWE-Agent with Claude Opus 4.7) achieves only a 68.3% success rate on AARRI-Bench.

Strict matter completion stalls (does not improve) despite stronger models.

Even frontier agents remain far from completing matters in a single pass.

Human-only teams take longer to complete the escape room than mixed human–AI teams.

Cellular research and development (R&D) is throttled by six structural processes that each consume months of manual engineering work per iteration: (i) synthesizing new features from standards or research papers into production code; (ii) conformance and interoperability testing; (iii) hardening against field anomalies and diverse deployment environments; (iv) data-driven optimization of network functionalities; (v) discovering and prototyping novel waveforms, functionalities, and capabilities for future standards; and (vi) securing the stack against vulnerabilities.

Mean-based metrics (e.g., tasks completed per worker-hour or mean handle time) can misrepresent AI's effects in workflows where tasks accumulate and compete for scarce human attention.

There is a 'speedup illusion' where people have accurate forecasts of independent completion times but significantly underestimate AI-assisted times.

More persuasive narratives may have had a detrimental effect on decision response times.

However, models remain limited in long-horizon reliability and domain-specific planning.

Extensive evaluations reveal that existing agents achieve only 36.0% task success on realistic media editing tasks.

Frontier agents struggle with end-to-end completion despite partial progress.

AI coding assistants expand the volume of code requiring review, turning code review into a growing bottleneck.

AI deployment reduces average chat duration.

This overthinking behavior significantly increases inference latency and energy consumption, forming a potential vector for denial-of-service (DoS)-style resource exhaustion.

Large reasoning models (LRMs) exhibit a tendency to "overthink", producing excessively long and redundant reasoning traces when confronted with incomplete or logically inconsistent inputs.

Participants using LLMs had significantly shorter idea-generation periods (p=0.0004).

Breaking down user stories into actionable tasks is a critical yet time-consuming process in agile software development.

Telemetry across 10,000+ developers shows 91% longer code review times.

There is a persistent female disadvantage in work intensity.

Standard PayGo degrades substantially under classroom-scale concurrency.

Each student query triggers several concurrent API calls whose latencies compound through a parallel-phase maximum effect that single-agent systems do not face.

Current AI agent frameworks have made remarkable progress in automating individual tasks, yet all existing systems serve a single user.

In an observational study of documented interactions across four AI tools (Claude, ChatGPT, Cowork, Codex), incomplete context was associated with 72% of iteration cycles.

People are more likely to give up after interacting with AI (increased likelihood of quitting tasks unassisted).

Resolution margin: the probability that posted queries are resolved declines because AI raises contributors' outside options, thinning the contributor pool and creating congestion on the platform.

Real estate pro forma development remains one of the most time-intensive functions in property investment, typically requiring twenty to forty hours per multifamily project through manual research, Excel-based modeling, and iterative scenario analysis.

Reliance on massive, schema-heavy prompts results in prohibitive per-token API costs and high latency, hindering scalable production deployment.

Using C.A.P. entails trade-offs: potential increases in latency and compute cost and a risk of over-correction (unnecessary clarification).

On-Premise RAG incurs higher latency compared with cloud RAG.

Integration cost: AI-generated outputs often require human revision, testing, and manual integration into existing systems.

We conducted a small-scale randomized experiment to measure uplift from human-agent collaboration on real-world computational reproducibility tasks.

The divergence between mean task speed and system-level delay caused by AI assistance is labeled the 'variance wedge'.

LLM guidance did not increase the total number of victims saved (no increase in total victims saved relative to baseline).

Larger-scale GPU workload results are projections calibrated from published benchmarks.

AI-assisted feedback does not reduce time per character (i.e., it does not increase time cost per unit of feedback).

The paper evaluates the proposed architecture using the outcome metric 'time-to-insight'.

The paper evaluates the proposed architecture using the outcome metric 'time-to-find'.