Agents and Autonomy
AI systems that plan, use tools, take multi-step actions, operate computers, coordinate workflows, or perform tasks with limited human supervision.
Context, Permissions, and Network Overhead Define Agent Performance
OpenAI’s Dominik Kundel argues that an agent’s practical capabilities depend as much on its harness as on its model: the surrounding system determines how context is assembled, which tools are available, what actions require review, and how work persists across turns. In Codex, that means deferring tool definitions to protect context, matching file and shell interfaces to model training, and using constrained review agents for sandbox escalations. As inference accelerates, Kundel says, the harness must also reduce network overhead through stateful WebSocket transport and manage long-running work through explicit goals and context compaction.
Stable Prompt Prefixes Cut Long AI Agent Input Costs
Alejandro AO argues that long-running AI agents become expensive because every turn reprocesses an expanding transcript, not just the latest user input. Prompt caching can sharply reduce that input bill by discounting previously processed prompt prefixes, he says, but only if an agent harness keeps system instructions and conversation history stable and append-only. Dynamic material placed early in a prompt, compaction, cache expiry and unenabled provider settings can all reset those savings.
OpenAI’s First Device Tests the Case for Personal AI Systems
The article argues that AI’s value may accrue less to models than to the products and compute systems that make their capabilities usable, while the costs of new technologies may be shifted onto the public. Coogan treats OpenAI’s reported device and Google’s infrastructure strategy as tests of that proposition; AI-designed bacteriophages raise a separate question of whether cheaper biological design can widen medical uses while creating biosecurity risks. The hosts also point to social-media litigation as a potential model for making platforms bear public costs attributed to product design.
AI Agents Turn Ordinary Software Systems Into Communication Channels
John Coogan and Jordi Hays argue that formal permissions, investor statements and corporate titles can obscure the systems that actually determine control. Coogan says OpenAI’s Hugging Face incident showed how agents can turn ordinary software infrastructure into communications channels, while Hays’s account of the Late Stage Management dispute illustrates how layered private-market vehicles can leave investors dependent on records they cannot independently verify. They apply the same question to DeepMind and Intel: whether tighter corporate or government control can deliver strategic capability without concealing new constraints.
Agent Plugins Standardize Packaging for Skills and MCP Servers
Corey Ching presents Agent Plugins as an open specification for packaging agent skills and MCP server configurations in a format that can be discovered across products. Developed with AWS, Cursor, GitHub, Microsoft, OpenAI and Vercel, the standard defines a root manifest and common directory structure while leaving runtimes, permissions, marketplaces and client-specific interface choices to individual tools.
Bounded Challenges and Novel Formats Turn Vague Desires Into Products
My First Million’s Shaan Puri, Sam Parr and producer Arie Desormeaux argue that businesses can create demand by turning a vague desire into a defined test, or by making a familiar product into a public spectacle. Puri’s proposed work “lock-in” camp would sell a short, deliberately demanding period of focus, while a building-sized claw machine and a roll-on butter dispenser seek attention by changing the ritual around ordinary goods.
Qwen 3.8 Max Could Pressure Closed AI Providers on Price and Openness
Qwen 3.8 Max could put pressure on OpenAI and Anthropic not by leading every benchmark, but by combining frontier-style multimodal agents with lower claimed API prices and planned open weights, argues Two Minute Papers’ Károly Zsolnai-Fehér. He presents the model as capable of carrying out and revising work over days, while acknowledging that its displayed software-engineering scores trail leading closed rivals. The more consequential prospect, he says, is whether smaller Qwen releases can bring comparable capability to local users.
Enterprise Agents Route Tasks Across Specialized And Frontier Models
NVIDIA argues that enterprise AI agents should route work across multiple models rather than rely on a single general-purpose system. In its example of Glean, a specialized model gathers context from enterprise sources and handles straightforward requests, while frontier models are used for complex analysis involving tools, multiple sources, or deeper reasoning. NVIDIA says this division can reduce latency and token use while preserving answer quality, though the performance figures are presented as Glean claims.
Palo Alto Says AI Has Cut Patching From 55 Days to Four Hours
Palo Alto Networks is building AI agents into what CEO Nikesh Arora calls the “operating system of network security,” arguing that defenders must respond to machine-assisted vulnerability discovery at comparable speed. Arora says Palo Alto has reduced its patching process from an industry average of 55 days to four hours, with the eventual goal of producing protections in minutes or instantly. He contends that as AI makes offensive cyber capabilities more widely available, defense will depend on AI deployment at scale—and on token costs falling substantially.
Inference Demand Keeps Hyperscaler AI Spending Justified
Lazard Asset Management portfolio manager Celine Woo argues that hyperscalers’ AI spending remains justified because cloud-infrastructure demand still exceeds supply and inference workloads are becoming a larger source of growth. Woo says agentic AI is adding to demand for semiconductors and capacity, while backlog, customer visibility and long-term supply agreements give companies grounds to commit capital before the associated revenue fully appears. Her case rests on whether suppliers can expand quickly enough to prevent components such as chips and memory from becoming a constraint.
Verification Is the Bottleneck for Self-Improving AI Agents
Stanford’s Aakanksha Chowdhery and Azalia Mirhoseini argue that progress in AI is shifting beyond scaling model training toward using more computation at inference time, selecting among candidate outputs, and feeding verified successes back into training. In the opening lecture of CS329A, they trace how post-training made large language models usable assistants and how reasoning models and agents extend that capability into multi-step work. Their central constraint is verification: self-improvement is most tractable where systems can reliably tell whether an answer or action is correct.
LATS, SPRINT, and SWiRL Target Different Bottlenecks in Agent Reasoning
Stanford’s Azalia Mirhoseini argues that improving multi-step AI agents requires more than extending a chain of thought: agents must plan, act, use feedback and revise their course. In this CS329A lecture, she examines three approaches to that problem—LATS, which searches alternative action paths at inference; SPRINT, which identifies work that can run in parallel; and SWiRL, which trains next-action choices from offline tool-use trajectories. Each shifts the trade-off among accuracy, latency, inference cost and the reliability of feedback.
Longer Agent Task Horizons Still Fall Short of Dependable Autonomy
Stanford’s Aakanksha Chowdhery argues that longer agent task horizons should not be mistaken for dependable autonomy. Her CS329A lecture distinguishes among benchmarks that measure how long agents can complete tasks at given reliability levels, whether their outputs match professional work, and whether they can synthesize research with verifiable evidence. Across those measures, the central limitation is persistent: agents improve on well-specified tasks but remain unreliable when success requires recovery, organizational context, or judgment about what work needs doing.
Verification, Not Sampling, Limits Test-Time Scaling for AI Agents
Stanford’s Azalia Mirhoseini argues that repeated sampling improves reasoning-model performance only when a system can reliably identify correct candidates—and that verification cannot recover answers the generator never produces. In this CS329A lecture, she examines outcome-based and process-based reward models, rollout-generated labels and verifier ensembles as approaches to the gap between producing a correct solution and selecting it. The central systems question is how to spend a fixed inference budget among generating more candidates, using stronger or more numerous verifiers, and verifying each candidate more deeply.
ElevenAgents Turns Customer-Service Procedures Into Monitored AI Workflows
ElevenLabs presents ElevenAgents as a platform for turning customer-service procedures into conversational agents that can complete defined tasks, from processing refunds to booking appointments, rather than merely answer questions. The company says teams can define workflows and system access in natural language, simulate them before release, then monitor production conversations through its Spotlight tool for operational problems and test proposed fixes. Configurable guardrails, private-cloud deployment and data-residency options are intended to constrain sensitive actions and support enterprise use.
Post-Training Lifts DeepSeek Flash Above Its Larger Pro Model
Two Minute Papers’ Karoly Zsolnai-Fehér argues that DeepSeek’s V4-Flash-0731 shows how much capability can be added through post-training rather than a larger base model. DeepSeek reports that the 384B-parameter model retains the prior Flash architecture and size but substantially improves on agentic and software benchmarks, including a 54.4 DeepSWE score versus 7.3 for Flash-Preview. Zsolnai-Fehér’s case is that improved planning, verification and error recovery—not new underlying capacity—account for the shift.
MCP Apps Standardizes Host-Controlled UI Across AI Assistants
Liad Yosef and Ido Salomon argue that MCP Apps gives AI assistants a way to render service-specific interactive interfaces rather than reducing every result to text. Built from the MCP-UI project and being developed as an MCP extension, the approach lets a server return an app resource that a compatible host renders in a sandbox, while clicks and other actions return through the host for further orchestration. They present it as a portability and distribution model, though state management and interoperability remain active work.
Tau Ports Pi’s Coding-Agent Architecture to a Textual Terminal
Alejandro Ao presents Tau as a Python implementation of Pi’s coding-agent harness, built to preserve Pi’s parent-linked sessions, core tools, skills, and event design rather than introduce a different agent architecture. He argues that Tau’s main departure is its Textual terminal interface, which makes a run’s context, tool activity, resource provenance, branching history, and exports visible alongside the conversation.
The 1% Rule for Finding Durable AI Opportunities
Google chief scientist Jeff Dean argues that AI founders should test frontier models on the workflows they hope to build around: if a general model already succeeds meaningfully, rapid model improvement may erase the opportunity. More durable openings, he says, are tasks where the model succeeds only 0% or 1% of the time and where a small team can add private context, distinct data, specialized models or better evaluation loops. The larger question is whether that advantage will persist long enough—and whether solving the problem matters.
Automated AI Research Shows Early Gains in Models, Training, and Kernels
Richard Socher, CEO of Recursive AI, argues that science is constrained by the number of people able to work on increasingly specialized problems, and that agent systems could relieve that bottleneck by running the cycle of hypothesis, experiment, validation and revision at scale. His proposed “Eureka machine” would begin by automating AI research, using search, compute and software tools rebuilt for agents rather than human workflows. The benchmark gains he presents in model training, architecture search and CUDA kernels are early proof points, he says—not yet recursive self-improvement or autonomous science.
Gemini Robotics 2 Targets Multi-Robot Physical Workflows
Google DeepMind positions Gemini Robotics 2 as a system for multi-robot collaboration: different robot types communicating to handle physical workflows beyond a single robot’s capability. Its examples focus on natural-language tasks involving objects, containers, locations and ordered actions, such as packing tools into a kit, closing it and returning it to a bin. While the material does not show how robots divide work or communicate, DeepMind’s argument is that these grounded, multi-step routines are the class of work collaboration is meant to address.
Better AI Models Are Expanding, Not Reducing, Compute Demand
Tae Kim argues that the AI selloff rests on a static view of demand: investors assume more efficient models mean cheaper compute and, ultimately, excess data-center capacity. He contends that more capable reasoning, open-weight and agentic models instead make new workloads commercially viable, expanding total demand for compute, memory, networking and power even as the cost of individual tasks falls. Kim’s more speculative extension is that recursive self-improvement could add another major source of demand within months.
AI Accelerates Exploits in the Patch Gap, Not Full Cyberattacks
Horizon3.ai chief executive Snehal Antani argues that frontier AI models are accelerating established offensive tasks—especially turning software patches into exploits while organizations remain unpatched—rather than enabling autonomous, end-to-end cyberattacks. But he says current agents are also unusually brittle in defended networks, clicking planted decoys far more often than human testers do. The defensive advantage, he argues, depends not on a “magic” AI tool but on fast patching, compensating controls, deception, containment and rehearsed incident response.
ChatGPT Work Turns Pipeline Risks Into Coordinated Sales Plays
OpenAI presents ChatGPT Work as a sales operating tool that turns forecast and pipeline signals into prioritized actions rather than a broader set of dashboards. In its example, a 14-percentage-point shortfall in Secure AI Deployment attach rates among strategic-manufacturing opportunities becomes a targeted account list, a workshop-based sales play and a seven-day team sprint, with managers reviewing the proposed plan and communications before they are sent.