VSvarunsingla.com

← All entries

Day 49· · 3 min read

Beyond LLM-as-Judge: Trajectory Evals, Continuous Production Monitoring & The New

Foundations & Protocols
Viral app of the day

STARS, #2 TRENDING) Originally launched October 2025, the Superpowers framework has now hit 121,000 GitHub stars and is

sitting at #2 on GitHub Trending as of May 2026 (40K stars in January, tripled in four months). It is a ode-based visual editor where you wire pre-built 'superpowers' (web browsing, file I/O, MCP tool calls, structured output) into autonomous agent workflows -- the visual equivalent of LangGraph nodes for on-developers. The viral hook is the 'skills' marketplace: community members publish reusable skill packs (web scraping, data cleaning, outreach), and any agent can install one with a click. It rides exactly the same 'skills as the new APIs' wave that powered Hermes Agent (47K stars in 8 weeks) and APM by Microsoft. Why it matters for evals: Superpowers ships built-in trace export to LangSm

VIRAL APP OF THE DAY -- SUPERPOWERS SKILLS FRAMEWORK (121K

Originally launched October 2025, the Superpowers framework has now hit 121,000 GitHub stars and is sitting at #2 on GitHub Trending as of May 2026 (40K stars in January, tripled in four months). It is a node-based visual editor where you wire pre-built 'superpowers' (web browsing, file I/O, MCP tool calls, structured output) into autonomous agent workflows -- the visual equivalent of LangGraph nodes for non-developers. The viral hook is the 'skills' marketplace: community members publish reusable skill packs (web scraping, data cleaning, outreach), and any agent can install one with a click. It rides exactly the same 'skills as the new APIs' wave that powered Hermes Agent (47K stars in 8 weeks) and APM by Microsoft. Why it matters for evals: Superpowers ships built-in trace export to LangSmith and Arize Phoenix -- every skill execution is an evaluable span by default. It is the first consumer-facing agent builder that treats observability as a feature, not a checkbox.

GitHub stars (as of May 2026) GitHub Trending stars added in 4 months

screener and more. Run as plugins in Claude Cowork/Code or as autonomous 'Claude Managed Agents' with full audit logs. Native Microsoft 365 add-ins (Excel, PowerPoint, Word, Outlook). New data partners: Moody's MCP server with credit data on 600M+ companies, plus Dun & Bradstreet, S&P, Verisk, Guidepoint. First credible agent template library for a regulated vertical.

four months. The Oct 2026 IPO anchor model (Mythos) plus Opus 4.7 + Cowork + finance agents are the

AI as a primary deployment surface for Claude in the enterprise.

Skills, now in ChatGPT and Codex for Plus/Pro/Business/Enterprise.

frontier-model testing by the US Center for AI Standards and Innovation. First formal government eval

of Annex III evidence for high-risk agent deployments.

Practical takeaways
PRACTICAL TAKEAWAYS Add trajectory evals this week. If you only score outputs to

PRACTICAL TAKEAWAYS Add trajectory evals this week. If you only score outputs today, you're missing 20-40% of failures. Pick LangSmith (LangGraph), Braintrust (eval-as-code) or Vertex AI (Gemini) and add trajectory_exact_match + trajectory_precision to your CI. Wire online evals on sampled prod traffic. Run a cheap LLM judge async on 5-10% of prod traces. This is your early-warning system for drift. Galileo, LangSmith and Braintrust all support this pattern with one config block. Build a golden dataset of 200-500 trajectories -- not outputs. Capture the full tool-call sequence, not just the final answer. Curate from real production traces, never synthetic. Promote new prod failures into it weekly -- the dataset self-heals. Treat Galileo Signals / LangSmith Insights Agent as your meta-eval. These auto-cluster failure modes from millions of traces. They find the bugs your evals miss. Even if you don't pay for them, the pattern (cluster traces by embedding → label clusters by LLM fi prescribe fixes) is easy to replicate in 200 lines of Python. Make every agent CI/CD pipeline a 5-stage gate. Lint → unit evals (trajectory + output) fi integration in Docker sandbox → shadow on prod traffic copy → canary with auto-rollback on online-eval drop. This is the standard for EU AI Act Annex III evidence by Aug 2 2026.

Tomorrow (Day 46): The Anthropic Finance Agents deep dive -- what each of the 10

Tomorrow (Day 46): The Anthropic Finance Agents deep dive -- what each of the 10 agents actually does, the Moody's MCP integration, the Microsoft 365 architecture, and what it means for the Big 4 / consulting model.

VS
Varun Singla
Singapore · About · Learning in public