{"slug": "using-multiple-models-to-extend-long-context-tasks-and-increase-build-viability", "title": "Using multiple models to extend long context tasks, and increase build viability", "summary": "A developer built a multi-model workflow that extends long-context tasks by using a frontier planner to refine user ideas into an append-only spec, a manager to decompose that spec into subtasks, and a build loop that hands off work between specialized models. The system stores layered knowledge including project context and prior decisions to guide each model, with the user managing all handoffs and able to halt any window. This architecture increases build viability by preventing context overflow and enabling each model to focus on its specific role within the larger task.", "body_md": "|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\nBuild Workflow\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nUser has Idea\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nUser + Frontier Planner

Refine idea, develop spec from idea.
Spec is append-only — core truth.
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nSpec (append-only)
Resolved view served
to Manager + Exit Mgr
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nLayered Knowledge
Project context, prior
decisions, anti-patterns
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\nBuild Loop\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nUser

Prompted for action or approval of spec change.
Manages all model handoffs.
Can halt any window.
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nFrontier Manager

Decompose spec into phased task list.
Audit gating at minor + major milestones.
Final call on retask vs halt — judgment, not rule.
Local fix → next task list. Structural → user prompt.
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nScoped Worker(s)

Executes task list, runs automated tests.
Stops at designated checkpoints.
Reports for verification by Auditor.
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nAuditor

Audits worker output vs codebase + spec.
Also acts as Manager context filter —
raw worker output never reaches Manager.
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nAuditor Output Contract
• verdict: pass / fail / partial
• one-line diagnosis
• rec action: continue / retask / escalate
• pointer: file / function / test
Manager pulls more detail only on demand.
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nFrontier Exit Manager
(was: Debugger)

User test program with debugger. Watches logs, applies surgical edits within bounded scope.
If structural fix needed → kicks back to Manager with diagnosis (re-enters audited build).
Needs whole-project context — sized for largest context window.
If pass → close session.
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nFinished App\" vertex=\"1\"> |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\ncontract\"> |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\nstructural fix\"> |\n|\n |\n|\n |\n|\npass\"> |\n|\n |\n|\n |\n|\nUser can halt any window\" vertex=\"1\"> |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\nFrontier (Planner / Manager / Exit Mgr)

Opus 4.6 / 4.7
GPT 5.4 / 5.5

Runner-ups
Gemini Pro 3.1
GLM 5.1
Kimi 2.6
Qwen 3.6 MAX

Exit Mgr: bias toward largest context window — needs whole-project view.
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nScoped Worker

DS Flash V4
Qwen 3.6 35B MoE (local)

Runner-ups
GLM 5.1
Kimi 2.6
Sonnet 4.6
Gemini Flash 3.0
Gemini Pro 3.1
GPT 5.3
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nAuditor

DS Flash V4
Qwen 3.6 35B MoE (local)

Runner-ups
GLM 5.1
Kimi 2.6
Sonnet 4.6
Gemini Flash 3.0
Gemini Pro 3.1
GPT 5.3

Same tier as Worker — context filter for Manager only works if Auditor is capable.
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nCLI / IDE layer
Opencode, Claude Code, VsCode, Ect
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\nLegend
— solid: data / control flow
— dashed: context read (spec / knowledge)
— red dashed: user halt / interrupt
\" vertex=\"1\"> |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |\n|\n |", "url": "https://wpnews.pro/news/using-multiple-models-to-extend-long-context-tasks-and-increase-build-viability", "canonical_source": "https://gist.github.com/tyoung515-svg/10d72830b3040b86f8e1687804a94cb0", "published_at": "2026-05-09 08:03:52+00:00", "updated_at": "2026-06-06 07:14:20.633690+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-agents", "ai-tools", "ai-research"], "entities": ["Frontier Planner"], "alternates": {"html": "https://wpnews.pro/news/using-multiple-models-to-extend-long-context-tasks-and-increase-build-viability", "markdown": "https://wpnews.pro/news/using-multiple-models-to-extend-long-context-tasks-and-increase-build-viability.md", "text": "https://wpnews.pro/news/using-multiple-models-to-extend-long-context-tasks-and-increase-build-viability.txt", "jsonld": "https://wpnews.pro/news/using-multiple-models-to-extend-long-context-tasks-and-increase-build-viability.jsonld"}}