Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchThe project described by Vitor Calvi is a collection of Jupyter notebooks for experiments in machine learning and local-first AI. The author says the notebooks are intended to run top-to-bottom in Google Colab or a local GPU runtime, with an emphasis on reproducibility, auditability and practical implementation—not leaderboard performance. That is the project’s stated aim, not an independently verified assessment of the notebooks or their current compatibility.
What the notebook lab is meant to do
Calvi presents my_public_notebooks as a way to make experiments easier to inspect and rerun. Each notebook is meant to answer two practical questions: “How do I run it without guessing?” and “What does this prove?” The author says notebooks should also make clear what they do not prove.
As an Amazon Associate I earn from qualifying purchases.
The stated rationale for open-sourcing the work is to make experiments more auditable, create shared baselines for local-first and on-device language-model work, and record failure modes that papers or READMEs may leave out. These are motivations, not measured outcomes.
What the collection covers
The article describes several distinct project areas rather than one unified experiment:
#1 Best Overall
- Auditable Recursive Latent Reasoner (RLR) experiments: described as using independently computed ground truth, checksums and held-out evaluation.
- Coconut and LFM2.5 production builds: described as continuous-thought approaches with KV-cache optimization and structured-decision workloads.
- Local MiroFish, Graphiti and Neo4j setup: Colab workflows described as having no external API dependencies.
- DSPark Swarms and API benchmarks: benchmark notebooks identified by the author.
- HRM product scenarios: scenario-focused notebooks.
- Bonsai27 Colab workflows: environment recipes and troubleshooting topics including ngrok tunneling and CUDA fixes.
These descriptions establish the intended scope, not the quality of results or present-day notebook status. The available article text does not independently verify the experiments, compatibility, repository license or current contents.
How to contribute usefully
The author invites contributions that improve reliability and make experiments easier to reproduce. Suggested work includes fixing broken cells, adding baselines, porting notebooks to CPU or MPS, adding production wrappers with timeouts, logging and error handling, documenting failure modes, and contributing reproducible notebooks.
Before opening a pull request
- Open an issue first to check whether someone is already working on the same change.
- Run the notebook from a fresh Colab runtime so the result does not depend on hidden state left by an earlier session.
- Keep the change focused, making it easier to review and reproduce.
- Remove secrets and private data; do not commit credentials or sensitive material.
- For a new notebook, explain what it demonstrates, what it does not demonstrate, and how to run it.
Where the author says help is needed
- LFM2.5 and Coconut: production wrappers, latency benchmarks and on-device execution paths.
- Graphiti and Neo4j local pipelines: alternative backends, memory persistence and evaluation harnesses.
- RLR: comparisons with Mamba-2, GRU-RSSM and transformer baselines under a common evaluation protocol.
- Colab environments: pinned-version recipes that document install order and T4 runtime pitfalls.
These are contribution priorities named in the article, not confirmation that each area is currently unfinished or that a particular implementation is available.
What readers should verify before relying on a notebook
The article describes the collection as public and runnable, but the available text does not identify a repository URL or establish its license, current commit, current notebook status or independently reproducible results. Readers should therefore treat its descriptions as the author’s account and verify access, dependencies and outputs in the notebook environment before using it as a baseline or building on its results.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
