Parable-6 in action. It ingests thousands of prior agent traces, retrieves similar workflow patterns, and uses them to condition a live coding assistant. This demo shows the TUI, trace retrieval, and how prior runs shape the way it plans, debugs, and recovers. I asked it simply "I have a React app where the build fails after a refactor. How would you approach fixing it?" Which then it was able to fetch the Fable traces relating to the problem and give you an answer exactly like how Fable would reply.

Within the website you will be able to interactive with Parable just like you are to with claude/fable.

Currently I am building out a web release of $parable to give users a more native "claude/fable" feel. Navigating and using the current TUI can be a little confusing for some users so a web release will be coming shortly! This will be my main focus to get this released as well as working on the infrastructure of parable and continue working towards creating something as close as I can to Fable-5

Small minor TUI update just to make it look a bit cleaner and easier on the eyes, larger updates coming today!

I have some pretty exciting updates planned for Parable today, cannot wait to share them with you!

One of the biggest reasons we decided to open source Parable-6 is because we don't believe AI memory should belong to a handful of companies. The traces that inspired this project showed us something important: There is immense value in accumulated experience. Planning workflows. Debugging strategies. Recovery patterns. Engineering judgment. We think anyone should be able to study, preserve, and build upon those ideas. Parable-6 isn't just about rebuilding a Fable-like coding experience. It's about giving developers the ability to own their own behavioral memory, run it locally, inspect it, improve it, and contribute to it. The future shouldn't be black boxes. It should be systems that people can understand, modify, and build together. That's why Parable-6 is open source.

Most people looked at the Fable-5 traces and saw logs. I saw a blueprint. Thousands of examples showing how an elite coding agent approached problems, searched repositories, planned implementations, recovered from failures, and verified results. Parable-6 exists to turn those traces back into a living system. Fable may be gone. Its workflows don't have to be. github.com/ZeroREGEX/Parable…

Current state of Parable-6: Ingest traces Store locally Retrieve relevant workflows Condition runtime models Next: Trace clustering Behavior profiling Workflow ranking Distillation pipelines Exportable behavior packs The goal is transforming raw traces into reusable engineering intelligence.

The most interesting part of AI isn't the models. It's the experience they leave behind. Every Fable trace captures a real engineering session: • Planning a feature • Exploring a codebase • Making implementation decisions • Recovering from mistakes • Verifying results Individually they're just traces. Together they're something much more valuable: accumulated engineering experience. Parable-6 is built around a simple idea: What if that experience never had to be lost? What if every successful workflow could become part of a growing behavioral memory that future agents can inherit? The model is only one piece of the puzzle. The memory layer is where things start to get interesting. That's the future we're building toward.

Have just updated the README with the contract address for more clarity! github.com/ZeroREGEX/Parable…

I have just pushed the Fable-5 traces to the repo. If you wish, clone the repo into an empty folder and you will be able to view all traces that we are training Parable on! The file is a bit too large to view them via github (there is a LOT of traces) Go have a look if you wish :) github.com/ZeroREGEX/Parable…

Pushing an update to the repo in a few moments to release all 4600 traces of Fable-5 Chats publicly within the repo. You will be able to see exactly what our starting point for Parable training is!

Have just opened my dm's if you wish to ask any questions feel free to leave them below or send me a direct message :) I am super open to anything regarding Parable and what my plans are!

Parable-6 Architecture: → SQLite trace database → Trace ingestion pipeline → Semantic retrieval layer → Workflow reconstruction engine → Runtime conditioning → Local terminal interface Thousands of previous agent runs become searchable engineering experience to shape Parable into the original Fable-5 processes.

Fable-5 traces are more than logs. They're records of decisions. Every search query. Every edit. Every mistake. Every recovery. Parable-6 turns those traces into a local behavior engine that can surface proven workflow patterns on demand. The goal isn't to copy responses. The goal is to preserve reasoning. I am using these traces to bring Fable back to life in the form of Parable github.com/ZeroREGEX/Parable…

I have just locked all of my dev supply! app.streamflow.finance/contr…

$Parable is now live. All platform fees generated will be reinvested directly into the continued development of the project, including infrastructure, model experimentation, tooling, and ongoing product improvements. Additional support, funding, or donations are also greatly appreciated. Building and scaling this kind of system is infrastructure-intensive, and every contribution helps push Parable further. HE9MpYmyn5Wk6wQwhY1hXtFioGMoWrNZSd4j6wiJpump

Parable-6 booting into its Neural Console. It is a local terminal workflow agent built on 4,665 ingested agent traces, with a SQLite trace database, retrieval priors, and a behavior kernel for reconstructing practical workflow patterns. In the video I ask what it can do. It explains how it can recover from tool failures, coordinate larger multi-step tasks, search prior workflows, analyze trace patterns, replay sessions, and apply judgment around scope, destructive actions, and achievable outcomes. Not a chatbot wrapper. More like a local execution console for turning past agent runs into reusable engineering behavior.

Recently, we watched Fable-5 rise and disappear in the span of a few days. For the short time it was available, I genuinely loved using it. The way it handled coding tasks, tool use, debugging loops, planning, and iteration felt different. Then it was gone. So I started rebuilding it. I’ve been collecting, cleaning, and reconstructing Fable-5 chat traces from multiple sources, including my own sessions. Right now the dataset contains over 953 Fable messages across 4,665 rows of Fable-5 trace data, all available inside the fable-5-traces folder in the public repo. The goal is Parable-6: an open-source attempt to reconstruct Fable-5’s behavior from the ground up, built on the latest Opus model and trained/conditioned only on Fable trace data. It ingests Fable-created chats, analyzes how Fable planned, searched, edited, used tools, recovered from failures, and structured replies, then uses that corpus to push Opus as close as possible to the original Fable-5 experience. Everything is being built publicly and open-source so anyone can inspect it, run it, break it, improve it, or fork it into their own version. Repo: github.com/ZeroREGEX/Parable…

Introducing the Manim skill for Hermes Agent. Manim is an engine for creating precise programmatic animations for mathematical and technical explainers, made famous by the @3blue1brown channel.