Agentic AI and autonomous systems
What it actually takes to put agents to work: the loops they run, the architectures that hold them, and where autonomy stops paying for itself.
Research sweeps
2026-08-25 · deep
Observability for Agent Orchestration, Pipelines and Swarms
Observability for LLM agent orchestration, pipelines and multi-agent swarms between August 2025 and August 2026: emerging standards (OpenTelemetry GenAI semantic conventions, MCP tool tracing), the tooling landscape (LangSmith, Langfuse, Arize Phoenix, Braintrust, W&B Weave, Comet Opik, Datadog LLM Observability), the two-layer pattern of durable orchestration plus agent tracing (Temporal, LangGraph), and how production operators such as Stripe, Shopify, Airbnb and Anthropic instrument, evaluate and debug agent fleets at scale
Claude Fable 5- financial
- frontier
- academic
- +3
2026-07-21 · deep
Agentic Harnesses Market Landscape, July 2025 to July 2026
Agentic harnesses from July 2025 to July 21, 2026: competitive landscape, market share by use case, model integration, orchestration patterns, and differences between coding harnesses and broader workflow agents, including Claude Code, OpenAI Codex, Cursor, GitHub Copilot, Google Jules, Replit Agent, Devin, Windsurf, OpenCode, Aider, OpenHands, LangGraph, CrewAI, AutoGen, and n8n.
GPT-5.6-sol- financial
- frontier
- academic
- +3
2026-05-10 · deep
Agentic RAG - Evolution, Challenges, and Decision Criteria
Agentic RAG between November 2025 and May 2026: how retrieval-augmented generation is shifting toward agent-driven architectures, the operational problems (token burn, context management, latency, reliability), information-organisation patterns such as context catalogues and semantic categorisation, parallels with traditional data warehousing (dimensions, measures, star schemas), the evolving RAG tooling landscape, and decision criteria for switching to pure agentic workflows.
Claude Opus 4.8- academic
- frontier
- tech
- +2
2026-04-30 · deep
Agentic Engineering And Enterprise Architecture Discipline
Agentic engineering after Andrej Karpathy's vibe coding meme, April 2025-April 2026: how AI coding agents are changing enterprise software engineering across security, testability, reliability, maintainability, availability, resilience, observability, operability, cost, recovery, and engineering governance.
GPT-5.5- frontier
- academic
- vc
- +3
2026-04-24 · deep
Engineering AI Control Plane
Engineering AI control planes for software delivery from July 1, 2025 through April 24, 2026: how teams implement AI across development workflows and CI/CD, choose tools/models/SDKs, govern observability and compliance, manage reliability and provider availability, and handle cognitive debt, dark code, case studies, success stories, and failure modes across team size, company scale, and greenfield versus brownfield systems
Claude Opus 4.8- financial
- frontier
- academic
- +3
2026-04-19 · shallow
The Karpathy Loop - AI Agents Running Autonomous Training Experiments
The "Karpathy loop" - autonomous AI agent research cycles that run and evaluate ML training experiments to discover improvements, April 2025–April 19 2026, including Karpathy's own explanations, independent commentary, and real-world implementations
Claude Opus 4.5- frontier
- blogs
- tech
2026-04-17 · deep
Agentic AI's Impact on Technology Operating Models and Architecture
Agentic AI's impact on enterprise technology operating models and architecture (January 2025–April 17th 2026): what stays (API infrastructure, data governance, SDLC controls), what shifts (DevOps as the new control plane, testing and rollback at agent speed, dark-code and agentic tech-debt governance), and whether frontier models like Anthropic's Mythos become embedded in CI/CD pipelines for security, code review, and release control
Claude Opus 4.8- financial
- frontier
- academic
- +3
2026-04-09 · standard
Enterprise Agentic AI Adoption Criteria
Enterprise agentic AI adoption in operational processes November 2025–present: procurement criteria, model drift risk, version stability, availability SLAs, and how enterprises manage dependency on AI vendors in production workflows
Claude Opus 4.8- financial
- frontier
- academic
- +1
Explainers
- Research Explainer · West (2026)
Free models agree with the expensive verifier when it says yes, but miss most of its real failures
Eight free candidates, five local open-weight models and three cloud free-tier routes, were retro-graded against frontier verdicts on ten real stages of agent-written code. The best caught 77% of genuine failures. Five caught almost none, while agreeing with passing work up to 98% of the time.
- Research Explainer · McCain (2026)
AI agents are running for longer, but good oversight is becoming more active
Anthropic's study of Claude Code sessions and public API tool calls finds growing practical autonomy, paired with a shift from approving every action towards monitoring, interruption and agent-initiated clarification.
- Research Explainer · Acharya (2026)
Governance maturity makes agent fleets safer, but the evidence is still simulated
A 750-run multi-agent simulation finds that higher governance maturity sharply reduces agent sprawl and risk incidents while improving task completion and composite Net Business Value. Its pivotal claim is that Level 3 is the minimum viable standard, not a decorative middle rung.
- Research Explainer · Lin et al. (2026)
An agent rewrites its own coding harness, and beats the engineers who used to do it by hand
Agentic Harness Engineering turns harness tuning into an automated loop. Ten iterations lift pass@1 on Terminal-Bench 2 from a bash-only 69.7% to 77.0%, past the human-built Codex harness and every self-evolving baseline.
- Research Explainer · Hitzig (2026)
Coding agents make a software background optional, but expertise still decides who succeeds
An analysis of roughly 400,000 Claude Code sessions finds a clear split of labour: people decide what to build, the agent decides how. Command of a domain, not the ability to write code, is what makes sessions succeed.
- Research Explainer · Liu et al. (2026)
Claude Code is a thin agent loop wrapped in a thick safety harness
A source-level reading of Anthropic's coding agent finds that about 1.6% of the code is AI decision logic. The other 98.4% is permission gates, context compaction, extensibility plumbing, and recovery.
- Research Explainer · Li (2025)
AI coding agents now ship 456,000 pull requests, but their code gets rejected far more often than human work
The first large-scale dataset of autonomous coding agent activity on GitHub reveals that speed and scale are real, but acceptance rates, review dynamics, and code complexity tell a more sobering story about the gap between benchmarks and production.
- Research Explainer · Fokou (2026)
Prompt guardrails can't protect AI agents that act on the world, so Parallax builds a wall between thinking and doing
A new security paradigm structurally prevents AI reasoning systems from executing actions, interposing an independent four-tier validator that blocks 98.9% of adversarial attacks with zero false positives, even when the agent is fully compromised.
- Research Explainer · Arunkumar (2026)
AI systems are evolving from text generators to autonomous agents, but the architecture for making them reliable is still being invented
A comprehensive survey proposes a six-dimension taxonomy for LLM-based agents, mapping the shift from simple reasoning loops to hierarchical multi-agent systems with standardized tool connectivity, and catalogues the open failure modes that still block real-world deployment.
- Research Explainer · Alenezi (2026)
AI agents are no longer just answering prompts, they're becoming goal-directed systems with their own control loops
This paper maps the architectural shift from stateless LLM calls to autonomous agent systems with typed tools, hierarchical memory, multi-agent coordination, and governance baked in from the start.
- Research Explainer · Okpala (2025)
AI agent crews can build and validate financial models, but they still need human oversight to stay safe
Researchers at Discover Financial Services built two collaborating multi-agent crews, one for modeling and one for model risk management, that autonomously handle the full ML pipeline on credit risk, fraud detection, and card approval datasets, matching or beating top Kaggle solutions while stress-testing their own outputs.
- Research Explainer · Humberd (2026)
Agency theory was built for human managers, but AI is becoming the agent nobody knows how to supervise
A new framework maps five stages of AI evolution against traditional agency mechanisms, arguing that firms need to scaffold monitoring and incentive systems now, well before AI gains full decision-making autonomy.
- Research Explainer · Gabison & Xian (2025)
LLM agents act on your behalf, but the law still holds you responsible when they fail
A principal-agent analysis of liability in LLM-based agentic systems reveals that delegation to AI agents creates legal exposure for users, providers, and platforms, with multiagent systems amplifying the problem far beyond what single-agent frameworks can handle.
- Research Explainer · Xia, Lu, Zhu et al. (2025)
Most LLM agent evaluation stops at launch; so the same failures keep recurring in production
A multivocal review of 161 sources reveals that academic evaluation overwhelmingly focuses on pre-deployment benchmarks. The authors propose EDDOps, a process model and reference architecture that make evaluation a continuous, governing function across the entire agent lifecycle.
- Research Explainer · Koc, Verre, Blank & Morgan (2025)
Your IDE should watch your AI's metrics; not just your code's syntax
A conceptual framework for wiring real-time LLM telemetry (traces, evaluations, prompt versions) directly into the code editor through the Model Context Protocol, turning prompt engineering from guesswork into a data-driven feedback loop.