01Log
Archive
The full log, newest first.
Fugu Ultra v2
Sakana’s high-performance multi-agent orchestration model, 1M context.
GPT Sol
OpenAI’s budget-tier GPT family alias with 1.1M context for cost-sensitive builders.
Nex N2.5 Pro
396.8B open-weights model under Apache-2.0, a self-hostable step up from the mini line.
OpenQuake Engine
OpenQuake Engine is open-source software for calculating and assessing seismic hazard, risk, and decision-making tools developed by the GEM Foundation. It is widely used, actively maintained, and substantial.
PySPH Framework
PySPH is a Python framework for Smoothed Particle Hydrodynamics simulations, with performance-critical parts in Cython and OpenCL. It is a real simulation engine that supports OpenMP, OpenCL, and MPI and is actively maintained.
FiPy PDE Solver
FiPy is a finite volume partial differential equation solver written in Python. It is a NIST-maintained research code that remains actively developed for solving PDEs.
Micromagnetic Simulator
MuMax3 is a GPU-accelerated micromagnetic simulator for magnetization dynamics. It is a domain-specific tool known for its performance and active development, with precompiled binaries and documentation available.
Multiphase Flow Solver
MFC is an exascale compressible multiphase flow solver in Fortran using Fypp metaprogramming. It ran the largest known public CFD simulation at 200 trillion grid points, scaling to over 43,000 GPUs and was a 2025 Gordon Bell Prize finalist.
Loopy Code Generator
Loopy is a transformation-based code generator for high-performance array-based CPU and GPU code. It is notable for enabling powerful compute optimizations and is actively maintained.
Gonum Numeric Libraries
Gonum is a set of numerical libraries for Go providing matrices, statistics, optimization, and more. It is a mature, widely used suite that is actively maintained with six-month releases.
Do Agents Know When They Succeed? Calibrating Agent Confidence from Internal Representations
Calibrates agent confidence from internal representations rather than final outputs — a step toward agents that know when they've actually succeeded and when to ask for help.
Subagents vs Agent Skills: Executing Reusable Knowledge for Long-Horizon Agentic Tasks
Head-to-head of subagents versus reusable 'skills' for long-horizon tasks — evidence on when spawning a subagent beats loading procedural knowledge, and vice versa.
AgentGrad: Intervention-guided Prompt Optimization for Multi Agent Systems
AgentGrad optimizes multi-agent prompts by sequentially intervening to identify which agent is failing, then clustering textual gradients to target fixes — a more surgical alternative to blanket prompt tuning across the whole system.
Muse, the band, lost its social media handles to Muse, Meta's new AI agent
Meta's AI agent 'Muse' took the band Muse's social handles — a 183-point HN thread on naming collisions between AI agents and existing brands.
GitHub Advanced Security
Expanded self-serve trials for Enterprise Cloud customers to test security tooling.
DeepSeek V4.1-Flash
DeepSeek V4.1-Flash announced — billed as smarter, faster, more efficient; check API docs for tool-call and context changes before swapping agent backends.
Devin CLI v3000.10.21
Devin CLI v3000.10.21 adds `/code`, `/smart`, and `/bypass` mode switches, per-tool disabling via `disabled_tools`, and a refusal fallback that retries on another model when a provider blocks a turn.
Antigravity Google — Gemini 3.8 Flash using ADC in AGY Enterprise
Antigravity Enterprise adds Gemini 3.8 Flash via Application Default Credentials — enterprise agents can use the Flash model with existing GCP auth.
Commerce Agent Blueprint
Anthropic's reference blueprint for building shopping and merchant agents with Claude, with runnable examples in retail, commerce, telecom, and entertainment, defining each agent once with prompts, skills, tool contracts, and gates.
Experiential Model Gateway
Experiential is an open-source gateway and router for agent workflows providing one OpenAI-compatible API for hosted, BYOK, and local models, with access controls and spend limits. It learns from production traffic to recommend models, train a specialized model you own.
HydraDB Graph Database
HydraDB is a distributed graph database written in Rust that stores data durably on S3-compatible object storage rather than local disks. It supports OpenCypher queries and Neo4j-compatible Bolt connectivity while keeping compute nodes stateless and easily scalable.
Gen1Recomp Pokemon Recreation
A native Lua/LÖVE2D recreation of Pokémon Red, Blue, Gold, Silver, and Crystal, with hand-written engine and map behavior and game data decoded from a player-supplied ROM. It is not affiliated with the impersonating gen1recomp[.]com website.
Claude of Duty
A browser-based first-person shooter built with Three.js and WebGL2, generated by AI agents from a single prompt, with roughly 55k lines across 11 subsystems and all textures, meshes, animations, and sounds procedurally generated at load time.
MiniMax H3 Engine
A native C inference engine for MiniMax H3 on Apple Silicon Macs, implementing prompt-to-video/audio, first/last-frame conditioning, and ordered references. It is built as a sequence of working vertical slices, with current work on Metal performance and memory optimization.
Agentic Video Editor
An open-source video editor built for agents, where edits become real code so agents like Codex, Claude Code, OpenCode, or Pi see the latest version. It handles cutting, filler removal, subtitles, color correction, animations, and rendering, with a headless mode.
Atria WHOOP Companion
Atria is an open-source, local-first iOS app and BLE research toolkit that reads compatible WHOOP straps over Bluetooth without a WHOOP account or cloud. It provides live heart rate, strain, sleep/workout evidence, and HealthKit export for privacy-focused strap owners.
AlterSend File Transfer
AlterSend sends files directly between devices over the internet with no cloud storage, accounts, or size limits. It uses peer-to-peer transfer with end-to-end encryption, though encryption details are not fully documented.
Neantik Antidetect Browser
NeAntik is an open-source, local-first antidetect browser for Apple Silicon Macs that runs a built-in Chromium with separate profiles, proxies, and Keychain storage. It requires no account, cloud, or telemetry and is intended for session isolation, privacy, development, and QA.
Bare Browser
Bare is a de-Googled Chromium build for Android that strips Google tracking, telemetry, and AI integration while retaining extensions and uBlock Origin. It adds background playback, media saving, and download manager choice, implemented as a patch series.
Relay Reverse Tethering
Relay shares an Android phone's internet connection with a Windows PC via reverse tethering over an encrypted WireGuard tunnel. It requires no root, server, or account.
PureSnitch Firewall
PureSnitch is a free, open-source macOS application firewall using pf-based blocking, DNS over HTTPS, and a SwiftUI rules manager. It is a signed, notarized Little Snitch alternative with no telemetry or subscription.
Personal Model Builder
Personal Model is an open-source, local-first long-term memory runtime that builds an evidence-linked HUMAN.md from Mac activity for Claude Code, Codex, and Cursor Agent. It gives AI tools inspectable context while keeping data private and user-controlled.
Nex N2.5 mini
35.1B text-generation model under Apache-2.0, a mid-size open option for builders who need capability without 700B overhead.
DeepSeek V4.1 Flash
763B-param image-text-to-text model under MIT; the open multimodal flagship builders can deploy without closed-API lock-in.
DeepSeek V4.1 Flash
A brand-new base model wearing a point-release number, not a fine-tune of V4.
GPT-Live-1
Natural voice experiences in the API, targeting real-time speech interactions for production voice agents.
tiny-aya base 32K
3.4B multilingual base model with 32K context under CC-BY-NC-4.0, a local option for long-context tasks.
GigaChat 3.5 432B A28B Reasoning
432B open reasoning model under MIT, aimed at builders who need explicit reasoning traces without proprietary APIs.