runtime

Jolt, a Lisp/Scheme language, recently added two features: program images in the style of Common Lisp and Smalltalk for full-state serialization and debugging, and a portable Scheme backend decoupled from the Chez runtime for architecture-agnostic portability.

Stanford researchers released Shepherd, a runtime layer that acts like Git for AI agent runs, recording typed events and enabling copy-on-write forks so agents can revert to previous states and reuse KV cache. Early tests show improved pass rates for multi-agent cooperative tasks.

Herdr, an open-source runtime and TUI for terminal coding agents, is joining Y Combinator's F26 batch. The project, which has 25k stars and 340k downloads, will remain Apache-2.0 licensed.

The author wrote a pure-MLX runtime for the vision and audio towers of Nvidia's Nemotron Omni model, enabling the full multimodal model to run locally on Apple Silicon Macs. All 23 tests pass against the PyTorch reference, and it achieves 67-152 tok/s on an M5 Max.

GraphARC is an open-source governed agent runtime that uses a deterministic checker to admit or refuse proposed agent graph actions before execution, providing audit trails and cost tracking. It demonstrates how to steer agents away from forbidden actions using structured refusals.

Cognition CEO Scott Wu will deliver a keynote at Runtime, discussing async background agents and the future of agent design.

Cloudflare released a preview library called computer that provides a SQLite-based virtual file system in Durable Objects, with support for multiple runtime backends (container, isolated shell, isolated JavaScript), offering persistent workspaces for Agents.

A detailed comparison of three local AI runtimes—Ollama, LM Studio, and llama.cpp—helping practitioners choose the right one for their workflow based on interface, API compatibility, quantization control, and more.

A new research architecture for AI agents is introduced, centered on a transparent runtime where every interaction becomes a replayable execution trail with full inspectability, including plan, execute, observe, verify, and remember stages.

elizaOS is an open-source, local-first operating system and runtime for building AI agents, featuring a model-agnostic runtime, plugin architecture, and multi-surface assistant capabilities.

Krasis, a MoE-focused runtime, enables running the 397B-parameter Ornith model on a single RTX PRO 6000 Blackwell 96GB GPU with ~20-24 tok/s decode by dynamically managing expert residency in VRAM.

Go 1.26 makes Green Tea the default garbage collector, improving cache-friendliness. This article visualizes heap allocation with Go and C#, and discusses challenges with non-moving collectors and sparse pages.

Go 1.24 replaces the old bucket-based map implementation with a Swiss Table-inspired design, improving cache locality, reducing pointer chasing, and boosting performance in many workloads.

A new runtime for the K and Q programming languages that reimagines execution using SIMD, parallelism, fusion, and compression to better leverage modern hardware.

The author built a durable runtime for orchestrating agent swarms, enabling reliable coordination of multiple AI agents.

Wasmtime 47 enables Wasm GC and exceptions proposals by default, making it easier for high-level languages with garbage collection and exceptions to target WebAssembly efficiently.

ActiveGraph shipped seven runtime releases (1.2.0 to 1.7.1) to harden its reference packs for autonomous agents, adding a safe fork-test-promote loop with subprocess isolation, manifest hashing, and complete event logging.

Ant is a JavaScript runtime and ecosystem including an engine, package manager, registry (ants.land), deployment platform, and desktop app framework (Ant Desktop) designed as a coherent alternative to existing JavaScript stacks.

The article benchmarks the WebAssembly-to-C approach using wasm2c with wide arithmetic support, demonstrating that it remains competitive with dedicated WebAssembly runtimes like Wasmer and Wasmtime in both speed and memory usage.

A deep dive into Go's profiling mechanism, explaining how the runtime collects CPU, heap, block, mutex, and goroutine profiles, and how they are represented in the pprof format.