01Log
Archive
The full log, newest first.
Show HN: JevBench, a reproducible benchmark for typed decision models
Show HN: JevBench — a reproducible benchmark for typed decision models, built to separate serious open-source implementations from lookalikes. 54 points.
Don't Mask the Environment: Observation Supervision Changes How Agents Explore Under RL
Supervising observations, not just actions, measurably changes how agents explore under RL — a counterintuitive lever for anyone tuning agent reinforcement learning.
Qwen3.8 Max Prime
Higher-throughput SKU of Qwen3.8 Max accepting text, image, and video at a premium price point.
Space Bunny Alpha
Anonymous large model with fast inference, strong coding, and native multimodal input — currently free.
GLM 5.3 Prime
High-throughput variant of GLM-5.3 with 1.5–2× faster output via inference acceleration, same 1M context.
Recognition, Simulation, and Refusal: A Contamination-Aware Study of Classic Psychological Effects in LLM Agents
Contamination-aware study of classic psychological effects in LLM agents — checks whether agents actually exhibit effects or just parrot contaminated benchmarks. Useful for anyone evaluating agent behavior.
Memory That Looks Forward: A Zero-Inference Prospective Term for Personal Memory Retrieval
Proposes a zero-inference prospective term for personal memory retrieval — memory that anticipates future queries without extra inference cost. Relevant to agent memory design.
RRSI: Regularized Recursive Self-Improvement of Agent Harnesses
RRSI formalizes regularized recursive self-improvement for agent harnesses — adds a regularization term to keep self-modifying loops from drifting. Directly relevant to anyone running agents that rewrite their own prompts or tools.
onPanda: Efficient Annotation of On-Policy Alignment Data for LLMs and Agents via Token-Level Correction
Open-sourced onPanda — token-level correction for on-policy alignment data, aimed at cheaper RLHF/RLVR annotation for LLMs and agents. Useful if you're building agent-specific alignment pipelines.
Gun Protocol
GUN is an open source protocol and ecosystem for syncing decentralized graph data, enabling encrypted, realtime peer-to-peer applications. It is notable as a novel p2p data store with CRDT-based synchronization, used in production by many apps.
Apache Cassandra
Apache Cassandra is an open source transactional distributed database with linear scalability and proven fault-tolerance on commodity hardware or cloud infrastructure. It is notable as a proven distributed wide-column database with exceptional scalability and fault tolerance.
OceanBase Database
OceanBase is a distributed relational database developed by Ant Group, using Paxos for high availability and linear scalability. It is notable for enterprise-grade distributed HTAP with vector support and very large cluster capacity.
Database Migrate CLI
golang-migrate/migrate is a database migration tool available as a CLI and Go library, applying migrations from various sources in correct order. It is notable as a robust and widely used migration solution for Go projects.
YugabyteDB SQL
YugabyteDB is a PostgreSQL-compatible, cloud-native distributed SQL database for mission-critical applications. It combines relational transactional consistency with NoSQL scalability, targeting cloud-native OLTP workloads.
Tile38 Geospatial
Tile38 is an open source in-memory geolocation data store, spatial index, and realtime geofencing server supporting points, bounding boxes, Geohashes, and GeoJSON. It is notable for specialized real-time geospatial querying and geofencing.
Realm Swift
Realm Swift is a mobile database for iOS, macOS, tvOS, and watchOS, serving as a replacement for Core Data and SQLite. It is notable for strong engineering, efficient resource use, and offline persistence.
Jev-Mem: System-One-Controlled Agentic Memory for Efficient AI Agents
Introduces a System-One-controlled memory layer for agents — fast, reflexive memory decisions instead of deliberative retrieval. Relevant for latency-sensitive agent loops.
Harness-Zero: Harness Distillation via Agent-as-Harness
Proposes distilling agent harnesses into the model itself — treating the harness as a trainable component rather than external scaffolding. Worth a look for anyone building custom harnesses.
OpenHands 1.21.0
Request Docker execution workspaces from Docker servers — moves execution off the local daemon. Also fixes stale conversation history merges and Forgejo PR/issue listing.
Claude platform release notes — September 18, 2026
Compliance API local session endpoints now return transcripts of Claude in Chrome sessions (beta, Enterprise) — closes a gap for orgs auditing agent activity in the browser.
Devin CLI v3000.11.1 — Devin Cloud in your terminal
Run Devin Cloud sessions from the CLI (`devin --cloud`), hand off/pick up PR branches between cloud and local (`/handoff`, `/pickup`), and SSH into cloud VMs (`devin ssh`). Also adds OpenTelemetry export and ACP model/thinking controls.
Grok 4.7
Flagship coding and agentic model succeeding Grok 4.6, with DeepSWE at 71%.
Ming Image 0.1 Design Layer
Open 6.2B image-text-to-image model for design-layer generation.
MiMo V2.6 Pro
Xiaomi's 1T+ parameter flagship with 1M context for demanding tasks.
MiMo V2.6 Flash
309B MoE with 15B active, 1M context at $0.14/$0.28 per 1M.
Antigravity Agent 09-2026
Google's updated Antigravity agent preview, replacing the 05-2026 build.
Portfolio Risk Management
Portfolio Construction and Risk Management book's Python code.
Actual Budget iOS
iOS client for Actual Budget
AI Marketing Skills
Open-source AI marketing skills — growth experiments, sales pipeline, content ops, outbound, SEO, and finance automation
Investor Skills Library
Investor Skills is an open-source library that turns durable investing judgment into portable, structured formats. It collects how great investors think, filter opportunities, size risk, and act under uncertainty — then structures those patterns so humans can study them and AI finance agents can apply them.
ICT MQ5 Library
ICT Concept OSS Library for MQ5/MT5
Market Research Dashboard
A single-screen real-time market dashboard for financial and industrial research: covering global indices (A-share/HK/US), commodities, treasury yields, sector hotspots, capital flows, 24/7 news, industry-chain watchlists, and AI token tracking.
Aurum Finance OS
Self-hosted personal finance OS — cash flow, net worth, budgets, and every income stream you own. Free for personal use.
BrowserKitten Web Agent
A Chrome MV3 unpacked extension that turns a logged-in browser into a programmable agent, letting users select elements on a live page, describe outcomes, and export editable office files. It runs locally with bring-your-own-key and no hosted server.
Navop All-in-One Workspace
A native all-in-one workspace for databases, SSH, SFTP, terminals, remote desktop, monitoring, and AI. It includes built-in support for MySQL, PostgreSQL, SQLite, DuckDB, SQL Server, Oracle, and ClickHouse, plus SQL editing, import/export, schema compare, and ER diagrams.
Vox Director Video
An agent skill that generates a finished Vox-style paper-collage explainer or ad video from a one-line topic, automating script, keyframes, motion, voice-over, music, and captions. Runs on Atlas Cloud plus ffmpeg, usable by coding agents.
Anything2Explainer Video
A Claude Code or Codex skill that turns any topic into a black-canvas motion-graphics explainer video with TTS voiceover, subtitles, and a chapter progress bar in Chinese or English. Every frame is drawn in code with Remotion, no stock footage.
DeepSeek IM Bridge
A plugin that connects IM bots to DeepSeek Harness via QR code, app manifest, or existing credentials across nine channels including Feishu, WeChat, DingTalk, Slack, Telegram, Discord, and WhatsApp. It manages multiple bots per channel from one settings entry.
Fluent Korean Plugin
A Claude Code output-style plugin that instructs Claude and coding agents to write clear, fluent Korean instead of broken machine-Korean with dropped particles and noun strings. It focuses on meaning over ornate style and can be adapted to other LLMs.
AI Slop Filter
A rule set for AI coding agents to reject generic AI-generated UI, text, and code. It includes 38 mandatory tiered rules and acts as a filter, not a style guide, while also auditing copy to avoid generic AI phrasing.