Live

▸Anthropic bets the next platform war is won in the agent runtime, not the model ▸The verifier is eating the generator: why test-time compute keeps winning on arXiv ▸Follow the money: AI funding is quietly rotating out of models and into the application layer ▸Your agent isn't unreliable — your evals are. Notes on the real production bottleneck ▸The small-model thesis finally pays off: where on-device AI actually makes sense in 2026 ▸Claude Opus 4.7 lands quietly, raises the floor for agentic coding ▸Postgres 18 made async I/O the default and nobody noticed ▸OpenAI's enterprise pivot is more interesting than its product pivot ▸Async traits in Rust are finally stable. We are going to regret this.▸Edge runtimes are out. Fluid compute is in. The post-mortem nobody asked for.▸Anthropic bets the next platform war is won in the agent runtime, not the model ▸The verifier is eating the generator: why test-time compute keeps winning on arXiv ▸Follow the money: AI funding is quietly rotating out of models and into the application layer ▸Your agent isn't unreliable — your evals are. Notes on the real production bottleneck ▸The small-model thesis finally pays off: where on-device AI actually makes sense in 2026 ▸Claude Opus 4.7 lands quietly, raises the floor for agentic coding ▸Postgres 18 made async I/O the default and nobody noticed ▸OpenAI's enterprise pivot is more interesting than its product pivot ▸Async traits in Rust are finally stable. We are going to regret this.▸Edge runtimes are out. Fluid compute is in. The post-mortem nobody asked for.

Issue · May 2026

The signal,
not the noise.

Glitch & Glow covers AI, data, dev and cloud with smart, occasionally deadpan analysis. Partly drafted by Claude, edited by humans.

Read the archive →How this is made

Articles: 0
Categories: 0
% Hype: 0%

Anthropic bets the next platform war is won in the agent runtime, not the model

MCP adoption crossed a threshold this quarter and the Claude Agent SDK quietly became the default scaffold for production agents. The model is no longer the moat — the runtime is.

May 28, 20262 min read

Latest

All articles →

The verifier is eating the generator: why test-time compute keeps winning on arXiv

A cluster of recent papers points the same direction: a smaller model that checks its own work beats a bigger model that doesn't. The economics of inference are quietly being rewritten.

2 min read

Follow the money: AI funding is quietly rotating out of models and into the application layer

The mega-rounds are still going to foundation labs, but the velocity has moved. Vertical, workflow-native AI companies are raising faster and at saner valuations. The application layer is where 2026 is being decided.

2 min read

Your agent isn't unreliable — your evals are. Notes on the real production bottleneck

The teams shipping agents that actually work in production share one habit, and it isn't prompt engineering. It's treating evaluation as a first-class engineering discipline with the same rigor as tests.

2 min read

The small-model thesis finally pays off: where on-device AI actually makes sense in 2026

Small models stopped being a compromise and started being a strategy. The economics, latency, and privacy math now favor running locally for a surprising share of real workloads.

2 min read

AI

Models, agents, capabilities and the discourse around them.

Anthropic bets the next platform war is won in the agent runtime, not the model

Yesterday

The small-model thesis finally pays off: where on-device AI actually makes sense in 2026

6d ago

Claude Opus 4.7 lands quietly, raises the floor for agentic coding

1w ago

Data

Pipelines, infra, analytics, governance.

Postgres 18 made async I/O the default and nobody noticed

2w ago

Dev

Languages, frameworks, the daily work of building.

Your agent isn't unreliable — your evals are. Notes on the real production bottleneck

4d ago

Async traits in Rust are finally stable. We are going to regret this.

2w ago

Cloud

Platforms, edge, infra economics.

Edge runtimes are out. Fluid compute is in. The post-mortem nobody asked for.

3w ago