Production AI agents work better when models propose and code decides
Odd_Huckleberry4363 · reddit · 2026-07-22
The poster describes a production pattern for AI agents: let the model propose actions, but keep deterministic code in charge of final commits and decisions.
In their job-search tool, the agent originally scored role matches mostly by judgment. That worked inconsistently, especially when hard constraints appeared later in the posting, such as:
- clearance requirements
- relocation or in-office expectations
- salary below the user’s minimum
- work authorization constraints
They redesigned the system so the model now handles only the fuzzy part of the score — roughly 40% — such as how well experience maps to the role and whether adjacent skills transfer. The rest is scored by deterministic rules for location, remote policy, compensation, clearance, work authorization, seniority, and other hard preferences.
Their takeaway: agents are more reliable when they propose and humans or code commit the final decision, especially for user-facing or budget-sensitive outcomes.
More from coding & agent
- A vibe-coded 10-part game for a girlfriend’s layover becomes a tiny AI love story — generativist · 2026-07-22
- X API and Grok Build can turn saved bookmarks into a personal intelligence system — XFreeze · 2026-07-22
- OAT uses 100 successful trajectories to debug failing AI agents without failure labels — TheTuringPost · 2026-07-22
- Azure Architecture Diagram Builder adds MCP support for agent-driven Bicep workflows — davemccollough · 2026-07-22
- Day 10 of a Fable 5 snowboarding game adds ski lifts, cable grinds and UI polish — FinanceYF5 · 2026-07-22
- AI Wayfinder demo says smart contracts and vibe-coded games can be built in minutes — templecrash · 2026-07-22