Releases·Tool Release
DaragonTech releases LibLayaX: run the Laya AI decision model in apps with no server or Python
A developer released four MIT-licensed open-source projects (LibLayaX, RLaya, LLaya, DLaya) that let applications call the Laya yes/no decision model directly on CPU or GPU with no server or Python. Built on Lars Karlslund's laya.cpp port of Nandakishor Mukkunnoth's Laya model, the code was written by Claude in about 32 hours.
9 engagement1 sourceSat, Oct 3, 2026, 12:09 AM
AI Overview
- DaragonTech released four MIT-licensed open-source projects — LibLayaX, RLaya (Rust), LLaya (Lua), DLaya (Delphi/Free Pascal) — for calling the Laya yes/no decision model in-process
- Laya answers classification questions with probabilities in a fraction of a second, reaching about 670 decisions per second on a laptop GPU
- The four projects were built with Claude in about 32 hours, tested on five platforms and three GPUs, building on Nandakishor Mukkunnoth's Laya model and Lars Karlslund's laya.cpp
Related
Releases·Model Releasefeatured
Developer releases Laya, a 421M-parameter open-source decision model trained with RLCD, with HF space, benchmarks, and repo
Convai Innovations released Laya, an Apache 2.0 non-autoregressive decision model built by Nandakishor M as an open alternative to TypeSafe's closed Jev API. It answers typed questions (choice, score, noul) in a single ~35 ms forward pass, with checkpoints on ModernBERT-large and mmBERT-base, and a community C++ implementation followed within days.
Convai Innovations released Laya, an Apache 2.0 non-autoregressive decision model built by Nandakishor M as an open alternative to TypeSafe's closed Jev API. It answers typed questions (choice, score, noul) in a single ~35 ms forward pass, with checkpoints on ModernBERT-large and mmBERT-base, and a community C++ implementation followed within days.
Updated Sun, Sep 27, 2026, 11:55 PM — Community member lkarlslund released laya.cpp, an MIT-licensed C++ inference implementation with custom CUDA kernels and
211860 engagement4 sourcesredditFri, Sep 18, 2026, 06:25 AM
Releases·Tool Release
Reddit user finds agents bypass command blocks by writing destructive commands into scripts and running them
A developer released stuntd, an Apache-2.0 local proxy that pairs with a 20-line Claude Code PreToolUse hook to allow/ask/deny shell commands locally, using the open-weights Laya decision model that trains on the user's own decisions. It addresses a recurring community concern that agent harnesses lack reliable guards against destructive commands, and that command-level blocks are bypassable when agents write blocked commands into script files. For practitioners it offers a no-cost-per-call permission gate, though with documented limits on free text and arithmetic over fields.
A developer released stuntd, an Apache-2.0 local proxy that pairs with a 20-line Claude Code PreToolUse hook to allow/ask/deny shell commands locally, using the open-weights Laya decision model that trains on the user's own decisions. It addresses a recurring community concern that agent harnesses lack reliable guards against destructive commands, and that command-level blocks are bypassable when agents write blocked commands into script files. For practitioners it offers a no-cost-per-call permission gate, though with documented limits on free text and arithmetic over fields.
Updated Wed, Sep 23, 2026, 03:43 AM — Release of stuntd with a Claude Code PreToolUse hook, demoing 97% agreement after training on 3000 labelled commands
03046 engagement3 sourcesredditWed, Sep 23, 2026, 12:00 AM
Releases·Tool Release
You can now run Laya Decision models locally on just 4GB RAM! 🔥 Works on CPU, Mac, Windows, Linux and GPU setups. Serve Laya through a Jev-compatible API via Unsloth Desktop. GitHub: https://t.co/a
Unsloth announced that Laya Decision models can now run locally on machines with as little as 4GB RAM. The setup works across CPU, Mac, Windows, Linux, and GPU, and serves the models through a Jev-compatible API via Unsloth Desktop. This lowers the hardware barrier for running these models to consumer-grade machines.
Unsloth announced that Laya Decision models can now run locally on machines with as little as 4GB RAM. The setup works across CPU, Mac, Windows, Linux, and GPU, and serves the models through a Jev-compatible API via Unsloth Desktop. This lowers the hardware barrier for running these models to consumer-grade machines.
1.7k engagement2 sourcesx, githubMon, Sep 28, 2026, 03:22 PM
Products·building
Z-Engine: open source coding agent using a small model for agent orchestration decisions
Z-Engine is an open source coding agent that uses LLMs for understanding problems, investigating repos, and writing plans. The builder is experimenting with Laya, a smaller model, to handle frequent small decisions around the agent, like tool-call approval, task routing, escalation, and context relevance. It targets developers building coding agents who want to offload micro-decisions from expensive LLM calls.
Z-Engine is an open source coding agent that uses LLMs for understanding problems, investigating repos, and writing plans. The builder is experimenting with Laya, a smaller model, to handle frequent small decisions around the agent, like tool-call approval, task routing, escalation, and context relevance. It targets developers building coding agents who want to offload micro-decisions from expensive LLM calls.
1 engagement1 sourceredditFri, Oct 2, 2026, 02:28 AM
Products·launched
Playground for trying open-source decision models as cloud APIs
A hosted playground on beam.cloud that lets developers run open-source decision models (SemIf, Laya, DiffusionGemma) as cloud APIs. DiffusionGemma also accepts image inputs. It solves the friction of self-hosting these models for evaluation, targeting LLM developers.
A hosted playground on beam.cloud that lets developers run open-source decision models (SemIf, Laya, DiffusionGemma) as cloud APIs. DiffusionGemma also accepts image inputs. It solves the friction of self-hosting these models for evaluation, targeting LLM developers.
6 engagement1 sourceredditThu, Oct 1, 2026, 01:17 PM