Week 81 — Agent foundations

LLM training, RAG, agents, and evaluation · Implement

Medium

Phase 7 — LLM training, RAG, agents, and evaluation

Overview

ReAct, plan-execute, tool calling, guardrails; agent harness (tool-approval gates, session persistence/resume, context compaction, sub-agent fan-out, sandboxing vs host-shell)

Mode this week: implement. Match the work to that mode (operating model): implement ships running code; read/diagram ships a doc; deploy/benchmark ships a measured run. Personal dates, checkboxes, and the RPG layer stay in a learner journal.

Core deliverable

Agent prototype

Gate 7 this week: RAG retrieval pipeline, retrieval evaluation report, prompt system, memory system, agent prototype

Optional depth

Add a production-shaped benchmark, cost or safety analysis, and an architecture trade-off note.

Week loop (from the cookiecutter journal)

Mon — Theory: ReAct, plan-execute, tool calling, guardrails; agent harness (tool-approval gates, session persistence/resume, context compaction, sub-agent fan-out, sandboxing vs host-shell)
Tue — Build: start the core deliverable above
Wed — Systems / papers: one paper or system note (see below if this week has assigned reading)
Thu — Eval / interview: write the four evidence items
Fri — Algorithmic Forge: Aho–Corasick tool/prompt detector or sequence-DP planner. (2-3 hr; Boss fight)

Forge theme: Week 81 - String/agent boss

Leetcode Darbar (parallel)

This week's band (70-81): Strings, tries, graphs, sequence DP. Production connection: Tokenization, retrieval, agents, guardrails.

Standard: one aligned problem. Auror: the five-slot queue on the Darbar page. Tracker stays in the learner journal.

Topics: graphs · string · trie

All Paths From Source To Target · medium · graphs
Binary Tree Paths · easy · string
Word Search Ii · hard · trie

Agent papers

This week sits on the GenAI agents track. Read the assigned paper and connect it to the focus above.

ReAct: Synergizing Reasoning and Acting in Language Models (arXiv 2210.03629) — Foundation of ReAct agents
LangGraph: Multi-Agent Workflows (LangChain Docs) — Graph-based orchestration, cycles, human-in-the-loop
Model Context Protocol (MCP) Specification (Model Context) — Standardized tool/data access for LLMs
AutoGen: Enabling Next-Gen Multi-Agent Applications (Microsoft Research) — Collaborative protocol design

Evidence contract

Every core week records all four (cookiecutter journal/weeks/week_081.md):

Code / implementation (or the design artifact on a writing week)
Benchmark / result
Design doc / technical explanation
Retrospective / learning note (Sailboat: destination, wind, anchor, rocks, heading)

On this site

Official Tensor-to-Tenant

Week 81 of Tensor-to-Tenant · Previous week · Next week.