Share

Hybrid By Design

Where Intelligence Earns Its Keep in Browser Automation

By Nora Kaplan— March 19, 2026

Feature image for article: Where Intelligence Earns Its Keep in Browser Automation

A browser automation workflow runs for the hundredth time. One open-source framework barely touches a language model. The other sends the full page state to one at every step, same as it did on run number one. Same task, same browser. Wildly different bills at the end of the month.

Both frameworks recently rewrote their architectures from the ground up. Both migrated to the same low-level browser protocol. Then they diverged on a question that has no obvious right answer, one with consequences reaching well beyond browsers into any workflow where intelligence could, in theory, touch every step.

Hybrid By Design

Where Intelligence Earns Its Keep in Browser Automation

By Nora Kaplan— March 19, 2026

A browser automation workflow runs for the hundredth time. One open-source framework barely touches a language model. The other sends the full page state to one at every step, same as it did on run number one. Same task, same browser. Wildly different bills at the end of the month.

Both frameworks recently rewrote their architectures from the ground up. Both migrated to the same low-level browser protocol. Then they diverged on a question that has no obvious right answer, one with consequences reaching well beyond browsers into any workflow where intelligence could, in theory, touch every step.

Learning By Doing

The Flight Simulator Problem

By Rina Takahashi— March 19, 2026

Feature image for article: The Flight Simulator Problem

More than 35 companies are now building practice environments for AI agents. Anthropic has discussed spending over a billion dollars on them. The pitch is intuitive: you wouldn't let a pilot fly without simulator time. Deeptune just raised $43 million from a16z on exactly this logic, building pixel-perfect replicas of Slack, Salesforce, and ticketing tools where models can rehearse. But flight simulators work because physics holds still. The live web does not. And there's a subtler problem inside these training environments, one that lives deeper than fidelity.

Learning By Doing

The Flight Simulator Problem

By Rina Takahashi— March 19, 2026

More than 35 companies are now building practice environments for AI agents. Anthropic has discussed spending over a billion dollars on them. The pitch is intuitive: you wouldn't let a pilot fly without simulator time. Deeptune just raised $43 million from a16z on exactly this logic, building pixel-perfect replicas of Slack, Salesforce, and ticketing tools where models can rehearse. But flight simulators work because physics holds still. The live web does not. And there's a subtler problem inside these training environments, one that lives deeper than fidelity.

Worth Reading

The 2026 MCP Roadmap

The protocol maintainers name the production friction points they haven't solved yet. That candor is rare.

Andreessen Horowitz Backs Deeptune's $43M Bet on Training Gyms for AI Agents

The Fortune exclusive that put RL simulation environments on the map as a standalone infrastructure category.

NVIDIA Launches Enterprise AI Agent Platform at GTC 2026

Migrate Your Semantic Kernel and AutoGen Projects to Microsoft Agent Framework

MCP vs A2A: Understanding How the Two Protocols Complement Each Other

The State of AI and Browser Automation in 2026

Worth Reading

The 2026 MCP RoadmapThe protocol maintainers name the production friction points they haven't solved yet. That candor is rare.

Andreessen Horowitz Backs Deeptune's $43M Bet on Training Gyms for AI AgentsThe Fortune exclusive that put RL simulation environments on the map as a standalone infrastructure category.

Quick links

NVIDIA Launches Enterprise AI Agent Platform at GTC 2026

Migrate Your Semantic Kernel and AutoGen Projects to Microsoft Agent Framework

MCP vs A2A: Understanding How the Two Protocols Complement Each Other

The State of AI and Browser Automation in 2026

Browser Agent Approaches

Browser automation used to be a framework choice. Pick Selenium or Playwright, write your selectors, move on. Now the interesting question sits one level up: should a browser be scripted, reasoned through by a model, or some combination that nobody fully agrees on yet?

Three distinct architectures have staked out genuinely different positions on where intelligence belongs in the automation stack. A fourth option, the one production teams keep arriving at on their own, suggests the answer is more pragmatic than any single philosophy. Here's how the landscape actually breaks down.

Browser Agent Approaches

Browser automation used to be a framework choice. Pick Selenium or Playwright, write your selectors, move on. Now the interesting question sits one level up: should a browser be scripted, reasoned through by a model, or some combination that nobody fully agrees on yet?

Three distinct architectures have staked out genuinely different positions on where intelligence belongs in the automation stack. A fourth option, the one production teams keep arriving at on their own, suggests the answer is more pragmatic than any single philosophy. Here's how the landscape actually breaks down.

AI Hybrid

Stagehand V3 Drops Playwright for Raw CDP

V3's big architectural bet: swapping Playwright internals for direct Chrome DevTools Protocol access, which buys a 44% speed gain on gnarly DOM structures like shadow roots and iframes. Element caching means repeatable actions skip the LLM entirely. AI only fires when the script can't cope.

Agent First

Browser Use Lets the Model Drive Everything

At 78,000+ GitHub stars, Browser Use takes the opposite position: hand an LLM a goal and let it reason its way through. Every single step involves live inference, blending DOM parsing with screenshot analysis. No pre-programmed paths. It adapts on the fly, trading predictability for autonomy.

Deterministic Baseline

Playwright Remains What Everything Else References

No AI, no inference calls. Playwright controls Chromium, Firefox, and WebKit through explicit selectors and scripted actions. With 78,600+ stars and 45.1% QA adoption, it's the deterministic bedrock. Now AI-connectable through MCP, it's quietly becoming the stable layer agent frameworks orbit around.

Production Pattern

Teams Keep Landing on an Eighty-Twenty Split

The pattern production teams keep discovering independently: Playwright handles the roughly 80% of steps that are predictable, AI frameworks cover the 20% that require genuine understanding. Pure AI wobbles at scale. Pure scripting shatters when UIs shift. The hybrid works because both failure modes are real.

Where Capital Is Flowing

Close to half a billion dollars moved this week into agent infrastructure. Not into building smarter models. Into everything around them: how agents learn enterprise workflows, how they get secured, how their code gets verified, and how they slot into the processes companies actually run.

Four separate rounds, four different layers of the stack, and a shared conviction underneath all of them. The bottleneck to enterprise agent adoption has shifted. Capability is table stakes now. What's fundable is trust, governance, and the operational scaffolding that makes deployment survivable.

Gartner's projection puts a timeline on the urgency. The money is racing ahead of an adoption wave already in motion, and where it lands tells you which gaps investors think will matter most over the next twelve months.

Where Capital Is Flowing

Close to half a billion dollars moved this week into agent infrastructure. Not into building smarter models. Into everything around them: how agents learn enterprise workflows, how they get secured, how their code gets verified, and how they slot into the processes companies actually run.

Four separate rounds, four different layers of the stack, and a shared conviction underneath all of them. The bottleneck to enterprise agent adoption has shifted. Capability is table stakes now. What's fundable is trust, governance, and the operational scaffolding that makes deployment survivable.

Gartner's projection puts a timeline on the urgency. The money is racing ahead of an adoption wave already in motion, and where it lands tells you which gaps investors think will matter most over the next twelve months.

Training Infrastructure

Deeptune Raises $43M for Agent Training Gyms

Andreessen Horowitz leads Deeptune's Series A to build reinforcement learning environments that simulate enterprise workflows across tools like Slack and Salesforce. The signal: RL training is becoming its own infrastructure category, bridging the gap between raw model capability and agents that can actually do specific jobs.

Agent Security

Kai Exits Stealth With $125M Security Platform

Evolution Equity Partners leads Kai's combined seed and Series A for a cybersecurity platform built from scratch around agentic threats. The signal: security and governance tooling has graduated from "nice to have" to primary investment thesis, with AI-driven attacks demanding AI-native defense.

Code Verification

Axiom's $200M Bet on Verifiable AI Code

Menlo Ventures leads Axiom's Series A at a $1.6B valuation for systems that mathematically verify code using Lean proof language. The signal: as AI writes more production code, formal verification is shifting from optional quality assurance to compliance prerequisite. Trust, made auditable.

Vertical Orchestration

Oro Labs Lands $100M for Agentic Procurement

Brighton Park Capital and Goldman Sachs Growth Equity co-lead Oro Labs' Series C after 300% revenue growth. The signal: vertical agentic apps with measurable, workflow-specific ROI are proving out as the enterprise adoption path investors trust most. Goldman's involvement says something about who's paying attention.