Exploring Next
Full archive →Episodes on the research, tools, companies, and ideas shaping what comes next.
-
EP 1185 AgentsDev ToolsLaunchVercelVercel BoardV0State Management In Language ModelsAgentic LoopsTool Use And Function CallingStructured OutputDurable ExecutionBlog
Introducing Vercel Board
Vercel Board is an experimental primitive: a shared message board where agents coordinate tasks, exchange context, and work together across your tools.
Search:
Firecrawl Script:
Haiku 4.5 Voice:
OpenAI TTS
-
EP 1184 AgentsData InfraLaunchFirecrawlAlexandriaTool Use And Function CallingAgentic LoopsModel RoutingToken EconomicsNews
Firecrawl launches Alexandria to connect your AI agents to more than 550 specialized data sources
Firecrawl reports more than 1.5 million users and 150,000 companies using its services, including teams at 67 Fortune 100 companies.
Search:
Exa Script:
Haiku 4.5 Voice:
Cartesia TTS
-
EP 1183 New ModelsAgentsLaunchGoogleGemini 4 NeonTool Use And Function CallingAgentic LoopsContext WindowToken EconomicsPrompt InjectionChain Of ThoughtBlog
Introducing Gemini 4 Neon
Today, we’re introducing Gemini 4 Neon
Search:
Tavily Script:
GPT-5.1 Voice:
Inworld TTS 2
-
EP 1182 New ModelsInferenceLaunchMicrosoftMai Cyber 2 FlashMai Code 2Model RoutingToken EconomicsInference OptimizationAgentic LoopsBlog
Seven new world-class models at scalable price-performance | Microsoft AI
Microsoft launches seven specialized AI models claiming significant cost-performance advantages, with MAI Cyber 2 Flash reaching 92.3% task success versus GPT-6 Astra's 91.5% at half the reasoning budget, and MAI Code 2 matching competitor performance at substantially lower cost.
Search:
Tavily Script:
GPT-5.6 Luna Voice:
Speechify Simba 3.5
-
EP 1181 New ModelsEvalsLaunchLiquid AID1 2D1 2 UltraCalibrationStructured OutputToken EconomicsBlog
d1.2: State-of-the-Art Omnimodal Decision Models | Blog
Introducing Liquid AI’s d1.2 and d1.2-ultra, state-of-the-art omnimodal decision models with audio support, five new tasks, and industry-leading accuracy.
Search:
SerpAPI Script: Nemotron 3 Ultra 550B A55B Voice:
OpenAI TTS
-
EP 1180 AgentsDev ToolsGitHubCopilotVs CodeState Management In Language ModelsDurable ExecutionAgentic LoopsTool Use And Function CallingRetry Loops And Error RecoveryBlog
Building the Agent Host Protocol for sessions that continue across devices
Discover how we built the open Agent Host Protocol to keep your agent sessions accessible across editors, apps, phones, and the cloud.
Search:
You.com Script:
GPT-5.5 Voice:
Cartesia TTS
-
EP 1179 AgentsDev ToolsLangchainManaged Deep AgentsAgentic LoopsState Management In Language ModelsTool Use And Function CallingContext WindowDurable ExecutionGraph Based Memory RepresentationAppend Only LoggingBlog
What We’ve Learned Building Proactive Agents
Proactive agents work toward goals in the background and check in when they need you. What we've learned about memory and UX from building GTM Agent and Robin.
Search:
Jina Script:
Sonnet 4.6 Voice:
Rime Coda
-
EP 1178 Dev ToolsAgentsLaunchVercelAPI Docs
UI SDK
A library of behaviors for agents to build the next generation of user interfaces.
Search:
Firecrawl Script:
GPT-5.4 Voice: ElevenLabs v4
-
EP 1177 AgentsTrainingClaude OpusGeminiPrompt EngineeringState Management In Language ModelsAgentic LoopsTool Use And Function CallingGraph Based Memory RepresentationFeedback LoopsResearch Paper
Self-Improvement Memory Enables Continual Recursive Self-Improvement
Harness optimization is often studied with a fixed policy model and benchmark. When either changes, the optimizer must decide which parts of an evolved harness to reuse or revise, but the harness alone may not explain why earlier edits were made or how they performed. In this work, we study Continual Recursive Self-Improvement (CRSI), a setting for continuing existing harness-optimization methods across compatible benchmarks or policy models. We propose Self-Improvement Memory (SIM) for the
Search:
Exa Script:
Nemotron 3.5 Lightning 30B A3B Voice: ElevenLabs v4 Turbo
-
EP 1176 InferenceDev ToolsPartnershipHugging FaceOpenrouterModel RoutingRetry Loops And Error RecoveryToken EconomicsBlog
Hugging Face and OpenRouter: Reliability Gains and Control Trade-Offs in Language Model Routing
Hugging Face partners with OpenRouter for automatic provider failover on language models, improving reliability but centralizing billing and removing custom provider keys.
Search:
Tavily Script:
GPT-5.6 Terra Voice:
Inworld TTS 2
-
EP 1175 Dev ToolsAgentsLaunchClaudeClaude CodeTool Use And Function CallingAgentic LoopsDynamic Code ExecutionState Management In Language ModelsConstraint VerificationAPI Docs
Run tool calls from code with Code Mode - Claude Code Docs
Try Code Mode, an experimental feature in which Claude writes JavaScript that makes several tool calls in one step.
Search:
Tavily Script:
GPT-5.4 mini Voice:
Speechify Simba 3.5
-
EP 1174 AgentsDev ToolsAnthropicClaudeOpus 5 5Agentic LoopsState Management In Language ModelsTool Use And Function CallingDurable ExecutionAppend Only LoggingToken EconomicsLong Horizon TrainingBlog
Building always-on agents
Always-on agents run indefinitely, deciding how and when to act across many sessions. This guide walks through their core primitives and how they fit together.
Search:
SerpAPI Script: Nemotron 3.5 Super VL 120B A12B Voice:
OpenAI TTS
-
EP 1173 Model Behavior New ModelsDev ToolsOpenAIHugging FaceOpenrouterModel RoutingToken EconomicsContext WindowAgentic LoopsTool Use And Function CallingState Management In Language Models
Model Behavior: Week of November 16, 2026
Hugging Face and OpenRouter turn model sprawl into a routing layer; OpenAI's Astra and Sol, Google's Hydrogen, and Claude Managed Agents split the question of who controls which model gets called.
Search:
Tavily Script:
GPT-5.5 Voice:
Cartesia TTS
-
EP 1172 New ModelsInferenceLaunchOpenAIGpt 6 2 AstraGpt 6 2 SolToken EconomicsModel RoutingAgentic LoopsTool Use And Function CallingAPI Docs
Introducing GPT-6.2 Astra and Sol
Introducing GPT‑6.2 Astra and Sol, with better professional outputs, coding, computer use, and efficiency at lower cost across ChatGPT Work, Codex, and the API.
Search:
Jina Script:
GPT-5.6 Luna Voice:
Rime Coda
-
EP 1171 Agent ObservabilityDev ToolsDatadogOpenAIAgentic LoopsTool Use And Function CallingRetry Loops And Error RecoveryState Management In Language ModelsConstraint VerificationCalibrationBlog
Compare AI agent changes with Online Experiments in Agent Observability | Datadog
Learn how Datadog Online Experiments, available through Agent Observability, enables you to test AI agent changes against real production traffic so you can compare response quality, cost, latency, and errors.
Search:
Firecrawl Script:
GPT-5.1 Voice:
Inworld TTS 2
-
EP 1170 InferenceDev ToolsPartnershipHugging FaceOpenrouterModel RoutingRetry Loops And Error RecoveryKV CacheToken EconomicsGitHub
Hugging Face and OpenRouter join forces to make open-weight models easy to run
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Search:
Exa Script:
GPT-5.1 Voice:
Speechify Simba 3.5
-
EP 1169 AgentsDev ToolsActively AIReinforcement Learning From Human FeedbackAgentic LoopsTask DecompositionState Management In Language ModelsFeedback LoopsBlog
Goal Engineering
The future of agents is them taking on open ended goals and driving towards completion of those goals. Learn more.
Search:
Tavily Script:
GPT-5.1 Voice:
Inworld TTS 2
-
EP 1168 AI SafetyTrainingAnthropicMetrAI FuturesBlog
Q3.5 Timelines Update: Looks Like the Intelligence Explosion Will Happen in 2027 After All
We wanted to get this post out quickly given the urgency of the topic.
Search:
Bright Data Script: GPT-5.1 Voice:
Speechify Simba 3.5
-
EP 1167 AgentsDev ToolsGitHub CopilotAndrea GriffithsTool Use And Function CallingAgentic LoopsTask DecompositionBlog
Build once, run on any agent: A practical guide to agent skills
Agent skills capture repeatable engineering work in files teams can review and reuse, while tools and permissions remain specific to the host.
Search:
SerpAPI Script: GPT-5.1 Voice:
Cartesia TTS
-
EP 1166 AgentsAgent ObservabilityLaunchAnthropicClaudeClaude Managed AgentsAgentic LoopsTool Use And Function CallingState Management In Language ModelsRetry Loops And Error RecoveryDurable ExecutionAppend Only LoggingBlog
Operate Claude Managed Agents inside your security perimeter
Introducing self-hosted Claude Managed Agents in private beta. Now you can run agents in your own environment, whether that’s in your own cloud or on-premise.
Search:
You.com Script:
GPT-5.1 Voice: ElevenLabs v4 Turbo
-
EP 1165 New ModelsAgentsLaunchGoogleGemini 4 HydrogenGemini 4 CarbonModel RoutingAgentic LoopsTool Use And Function CallingToken EconomicsContext WindowInference OptimizationRetry Loops And Error RecoveryTask DecompositionBlog
Introducing Gemini 4 Hydrogen
Today, we’re introducing Gemini 4 Hydrogen, our daily driver for agentic workflows
Search:
Jina Script:
GPT-5.1 Voice:
Rime Coda
-
EP 1164 AI SafetyTrainingReinforcement Learning From Human FeedbackReward HackingGroup Relative Policy Optimization GrpoSampling And TemperatureEnsemble MethodsKl DivergenceResearch Paper
Theoretical Limits of Language Model Alignment
Language model (LM) alignment improves model outputs to reflect human preferences while preserving the capabilities of the base model. The most common alignment approaches are (i) reinforcement learning, which maximizes the expected reward under a KL-divergence constraint, and (ii) best-of-$N$ alignment, which selects the highest-reward output among $N$ independent samples. Despite their widespread use, the fundamental limits of reward improvement under a KL budget remain poorly understood. We
No Search Script:
GPT-5.6 Luna Voice: ElevenLabs v4
-
EP 1163 AgentsDev ToolsGoogleCursorAnthropicTask DecompositionAgentic LoopsTool Use And Function CallingState Management In Language ModelsConstraint VerificationBlog
Google /swarm vs Cursor Multiplayer: Who Do Your Agents Actually Work For?
Compares Google's /swarm agent coordination for individual developers against Cursor's multiplayer Projects that wire agents across team members, arguing the real shift is orchestration design rather than agent capability.
Search:
Exa Script:
GPT-5.4 mini Voice:
Inworld TTS 1.5 Mini
-
EP 1162 AgentsInferenceLaunchOctenArtificial AnalysisRetrieval Augmented GenerationAgentic LoopsRerankingEmbeddingsToken EconomicsNews
Ex-Alibaba engineer's new startup Octen pairs 10-millisecond AI search with a $20 unlimited plan
Octen’s research paper describes a technique called generative retrieval. Instead of searching a conventional index and repeatedly ranking candidate pages, a trained model produces an identifier that maps to an existing URL.
Search:
Tavily Script:
GPT-5.1 Voice:
Speechify Simba 3.5
-
EP 1160 AgentsEvalsPluraiClaudeChatgptSynthetic Data Generation For ValidationTask DecompositionFine TuningContext WindowDirected Acyclic GraphNews
Plurai wants to help AI agents work better using the same approach that helped get self-driving cars to roads | TechCrunch
Like the end-to-end models trained to drive cars, AI agents are probabilistic systems operating in unpredictable environments.
Search:
SerpAPI Script: GPT-5.1 Voice:
Inworld TTS 2
-
EP 1159 New ModelsInferenceLaunchTrustedrouterOpenAIAnthropicModel RoutingEnsemble MethodsToken EconomicsNews
TrustedRouter launches three new encrypted 'fusion' models that combine multiple LLMs, including one with exclusively U.S. models
In an exclusive video interview with VentureBeat, founder and CEO Joseph Perla explained how the models work and why the company wants to give developers more ways to match models to the work they’re doing.
Search:
You.com Script:
GPT-5.1 Voice:
Speechify Simba 3.5
-
EP 1158 AgentsDev ToolsGoogleAgentic LoopsTool Use And Function CallingContext WindowPrompt InjectionTask DecompositionBlog
Agent Distraction- Google Developers Blog
An empirical analysis of coding agent distraction from legacy inline TODOs, quantifying the overhead in tokens, execution steps, and developer wait time.
Search:
Jina Script:
GPT-5.1 Voice:
Cartesia TTS
-
EP 1157 AgentsAI SafetyBenchmarkRogueAgentic LoopsTool Use And Function CallingTask DecompositionRetry Loops And Error RecoveryPrompt EngineeringIn Context LearningConstraint VerificationResearch Paper
ROGUE: Evaluating Corrigibility Failures in Frontier Computer-Use Agents
As AI agents are increasingly deployed in real personal and corporate settings (email accounts, development workflows, company databases, etc.), safety considerations surrounding these agents become paramount. Although much work has focused on agent safety in the presence of an adversary, we study corrigibility: whether agents remain amenable to human correction, interruption, or shutdown while pursuing benign tasks. We introduce ROGUE, a benchmark in which agents are asked to complete
Search:
Firecrawl Script:
GPT-5.6 Luna Voice:
Rime Coda
-
EP 1156 Model Behavior AgentsInferenceMuse EmberInkling 2 SmallAxion 1 0Tool Use And Function CallingModel RoutingState Management In Language ModelsAgentic LoopsMixture Of ExpertsReinforcement Learning From Human FeedbackLong Horizon Training
Model Behavior: Week of November 9, 2026
Frontier labs chase unified giant models while Meta's Muse Ember, Inkling 2 Small, and ByteDance's Axion 1.0 standardize on planner-executor stacks—the real competition shifts to orchestration layers controlling how agents snap together.
Search:
Tavily Script:
GPT-5.1 Voice: ElevenLabs v4
-
EP 1155 Overview AgentsTrainingLongcat 2 5On Policy LearningReinforcement Learning From Human FeedbackReward HackingGroup Relative Policy Optimization Grpo
Overview: On-Policy Learning
A cyclist learns from wobbles on today's bike. On-policy learning means the system studies its own current behavior, then adjusts.
Search:
Tavily Script:
GPT-5.6 Terra Voice: ElevenLabs v4 Turbo
- No episodes match these filters.
Powered by
Cloudflare Twilio Claude
OpenAI