Anthropic moves Computer Use out of beta, ships native sandbox primitive
Claude's screen-grounded agent loop graduates with new tool-use primitives, an isolated sandbox, and tighter rate-limit policy for production deployments.
Research worth reading
Browse all →Research · Northeastern University
Reflexion, three years on: what self-critique still buys you
A meta-analysis of 41 papers building on Reflexion-style self-critique loops finds modest, durable gains in coding and tool-use, and diminishing returns in open-ended reasoning.
Research · Stanford NLP
Long-horizon memory: survey of seven architectures, ranked by recall and cost
Compares episodic, semantic, hybrid, and graph-based memory across realistic 30-day agent simulations. Hybrid stores win on recall; graph stores win on cost stability.
Research · DeepMind
Six failure modes in tool-using agents, and the patterns that fix them
An empirical taxonomy of agent tool-use failures across 4,000 traces from production deployments. Schema drift and silent partial-failure dominate.
Tools builders are using
All 20 tools →Arize Phoenix
Arize AI
OpenTelemetry-native LLM observability and evaluation.
AutoGen
Microsoft Research
Conversational multi-agent framework with strong reasoning patterns.
Browserbase
Browserbase
Hosted, isolated browsers for agent automation with session replay.
Continue
Continue
Open-source coding-agent IDE extension for VS Code and JetBrains.
CrewAI
CrewAI
Role-based multi-agent framework with declarative crew definitions.
E2B
E2B
Cloud sandboxes for code-running AI agents.
Haystack
deepset
Pipelines for retrieval-heavy agent workloads.
Helicone
Helicone
Lightweight LLM observability with a proxy-first model.
The Agent Brief
Three things in agentic AI, every Tuesday.
What changed, what matters, what builders should do next. No hype. No paid placement.
Build
All guides →- Ship a browser agent on Browserbase
A production browser-agent stack with anti-bot resilience, session replay, and a kill switch.
- Cost controls for agent workloads
Token budgets, fallback tiers, and the dashboards that catch runaway runs before they hurt.
- Build a replay-based eval set in a weekend
How to capture, redact, and score real production sessions to evaluate agent candidates.
- Build your first production agent with LangGraph
A 90-minute walkthrough that ships a tool-using agent with persistent state, retries, and observability.
- Add long-term memory with Letta
Wire a hierarchical memory store into an existing agent and audit what it remembers.
Use cases
All →- Customer support: agent-led deflection at the contact moment
How leading B2C teams are reducing tier-1 ticket volume by 35-55% with a tightly-scoped support agent.
- Engineering: internal-tool agents over the API graph
How platform teams replace one-off internal dashboards with a shared agent over their API graph.
- Legal: contract redlining assistant
A focused agent flags deviations from a playbook and proposes redlines for a human to approve.
- Sales: pre-call research and account briefs
A research agent assembles a 1-page brief 30 minutes before every external call.
- Security: alert triage and enrichment
An agent enriches and triages SOC alerts, halving the load on tier-1 analysts.