{"slug": "ai-builder-notes-week-of-june-14-2026", "title": "AI Builder Notes - Week of June 14, 2026", "summary": "A developer describes the 'loop' pattern for building reliable AI agents: goal, agent acts, verifier checks, state/memory updates, policy decides next action, repeat/stop/escalate. The developer also highlights Fable 5's exceptional capabilities, particularly in a table tennis spin demonstration, and notes that the U.S. government banned Fable shortly after its release following a reported jailbreak. OpenRouter launched Fusion, a council-of-LLMs feature, and Google released the Open Knowledge Format for curated reusable context.", "body_md": "My thoughts and my twitter’s feeds thoughts\n\nThis week was all about the ‘loop’ and [Fable](https://platform.claude.com/docs/en/about-claude/models/introducing-claude-fable-5-and-claude-mythos-5).\n\nThe best way I can describe it is: design the flowchart. Think of the deterministic flowchart on how you want your agents to work.\n\nAim to have:\n\nmore deterministic bits - this keeps things more predictable\n\nmore verification bits - this is agent feedback\n\nmore useful tool calls - tests, logs, screenshots, repo inspection etc. - this gives the agent feedback.\n\nThe ‘loop’ is essentially:\n\n`goal -> agent acts -> verifier checks -> state/memory updates -> policy decides next action -> repeat/stop/escalate`\n\nnow the specific implementation of this - will differ based on what you’re working on.\n\nIf you notice you are repeating a certain workflow manually - time to DAG it up.\n\nThe claude code dynamic workflows feature let the model write that DAG for you. That is fine for exploratory, reversible work.\n\nFor production software, the DAG is the product: you should write the stages, checks, stop conditions, retries, and review gates yourself.\n\nFable capabilities are absolutely insane, I tried it myself and it is entirely worth it for you to spend 2 minutes looking at this.\n\nThere are a few projects that I fire up a new model into to see what’s it gonna do.\n\nA project I wanted to build was a way to teach and demonstrate ‘spin’ in table tennis, every frontier model before Fable fumbled hard. But Fable outshined them with ease: [https://srijanshukla.com/artifacts/spin-lab/](https://srijanshukla.com/artifacts/spin-lab/)\n\nIf you personally did not experience a big shift in capability, you are probably not asking it a complex enough or ambitious enough task.\n\nFable came, and Fable was taken away. The United States Government(USG) was reported with a `jailbreak`\n\n- which Anthropic considers not significant. The USG anyway banned Fable just after few days of release. Big drama.\n\nFable was very pricey $$$$\n\nHence, people developed some patterns of work on those few golden days of Fable being available.\n\n- use Fable as planner/architect/taste/spatial/front-end judge.\n\n- use GPT-5.5/DeepSeek/Kimi as executor/worker.\n\nOpenRouter released their `Fusion`\n\nfeature as a model on their platform, accessible via API. Fusion is basically council-of-LLMs pattern - providing results that they claim can rival the frontier Fable 5 solo.\n\nGoogle Open Knowledge Format - [https://github.com/GoogleCloudPlatform/knowledge-catalog/blob/main/okf/SPEC.md](https://github.com/GoogleCloudPlatform/knowledge-catalog/blob/main/okf/SPEC.md) - the next iteration I think of LLMWiki.\n\nThis is **“curated reusable context”**\n\nI seem to have forgotten where I saved this from, but a great way to think about how much trust can be given out to your friendly neighbourhood model,", "url": "https://wpnews.pro/news/ai-builder-notes-week-of-june-14-2026", "canonical_source": "https://dev.to/srijanshukla18/ai-builder-notes-week-of-june-14-2026-2a6g", "published_at": "2026-06-15 03:45:18+00:00", "updated_at": "2026-06-15 04:11:00.421215+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-agents", "ai-policy", "ai-products"], "entities": ["Fable", "Anthropic", "OpenRouter", "Google", "GPT-5.5", "DeepSeek", "Kimi", "United States Government"], "alternates": {"html": "https://wpnews.pro/news/ai-builder-notes-week-of-june-14-2026", "markdown": "https://wpnews.pro/news/ai-builder-notes-week-of-june-14-2026.md", "text": "https://wpnews.pro/news/ai-builder-notes-week-of-june-14-2026.txt", "jsonld": "https://wpnews.pro/news/ai-builder-notes-week-of-june-14-2026.jsonld"}}