Lance Martin

This account is verified. Learn more

@RLanceMartin

MTS @anthropicai

San Francisco, CA rlancemartin.github.io Joined May 2009

493 Following 42K Followers 1.4K Posts

Latest posts 5

Lance Martin @RLanceMartin 1d

check out my new article in the Anthropic eng blog.

i talk about "dreaming" (offline compute) as a way to organize Claude's memory. it can correct mistakes, create memories, or build new representations.

x.com/AnthropicAI/st…

Anthropic Verified account @AnthropicAI 1d

Memory is what allows Claude to learn from experience. In a new Engineering Blog post, we share how “dreaming” helps Claude organize what it’s learned, correct its mistakes, and find what it needs more easily the next time. anthropic.com/research/compo…
View quoted post on X

Lance Martin @RLanceMartin Oct 8

i wrote a short guide on building automations with Claude Managed Agents.

it covers a few topics for agent deployments like credential vaults, guardrails, & memory.

@cjav_dev + i share a quickstart repo & a simple Claude Code command to set it up.

claude.dev/blog/building-…

claude.dev Building effective agent automations / claude.dev Blog A reference implementation walkthrough with common failure modes

Lance Martin @RLanceMartin Oct 7

Haiku 5.5 is 20x cheaper than Sonnet 5.5 (for <100k input tok), fast, and great for sub-agent parallelization.

i've been building maps w/ many Haiku 5.5 sub-agents that stream drawing commands to a shared canvas:

0:41

Lance Martin @RLanceMartin Sep 28

Claude Code can help with eval design + hillclimbing.

check out this article i just published for details:
claude.dev/blog/automatin…

claude.dev Automating eval design and hillclimbing with Claude / claude.dev Blog Principles for designing evals and hillclimbing against them without fooling yourself, and how the claude-api skill's build-eval and hillclimb commands put them to work.

Lance Martin @RLanceMartin Sep 28

i tested sonnet 5.5 on "code-to-painting". every pixel is generated by python the model wrote. it's shown a photo, writes a brush engine, renders, looks, and revises. you can see the large step-up from sonnet 5 in terms of coding + visual reasoning.

0:25