Nothing is editorialised. Sources land, the tape prints, every fact keeps its link back to where it came from. The archive is the product.
Artificial Analysis observed and listed GLM-5.3 (max) in its model catalog on 18 August 2026, scoring 60 on the AA Intelligence Index. This is an observation from an independent evaluator, not a vendor announcement — the model's creator label as AA lists it is 'Z AI'.
Artificial Analysis added G9v3-39A5B to its model catalog on 2026-08-19, measuring it at 34 on the AA Intelligence Index. The creator is listed as AI9Stars. This is an independent evaluation, not a vendor announcement.
Artificial Analysis added LFM2.5-2.6B to its model catalog on 2026-08-21, measuring it at 11 on the AA Intelligence Index. The creator is listed as Liquid AI. This is an independent evaluation, not a vendor announcement.
Artificial Analysis added Qwen3.8 27B (medium) to its model catalog on 2026-08-21, measuring it at 44 on the AA Intelligence Index. The creator is listed as Alibaba. This is an independent evaluation, not a vendor announcement.
Artificial Analysis added Qwen3.8 27B (low) to its model catalog on 2026-08-21, measuring it at 43 on the AA Intelligence Index. The creator is listed as Alibaba. This is an independent evaluation, not a vendor announcement.
Artificial Analysis added Grok 4.6 (xhigh) to its model catalog on 2026-08-21, measuring it at 60 on the AA Intelligence Index. The creator is listed as SpaceXAI. This is an independent evaluation, not a vendor announcement.
Artificial Analysis added Grok 4.6 (medium) to its model catalog on 2026-08-21, measuring it at 59 on the AA Intelligence Index. The creator is listed as SpaceXAI. This is an independent evaluation, not a vendor announcement.
Artificial Analysis added Grok 4.6 (low) to its model catalog on 2026-08-21, measuring it at 52 on the AA Intelligence Index. The creator is listed as SpaceXAI. This is an independent evaluation, not a vendor announcement.
Buzz Desktop adds a workflow editor and workflow reply-in-thread support, ships compact Buzz link metadata in messages, makes Projects workspaces selectable with context-aware collaboration, and closes several notification, agent provenance, and performance gaps.
The ID z-ai/glm-5.3 showed up in OpenRouter's public model-list API on 2026-08-21, listed at 1,048,576 context. Catalog observation only — no Z.ai announcement was checked, and the entry could be a test or placeholder.
A dated DeepSeek V4 Pro model ID appeared in OpenRouter's public model-list API on 2026-08-21, listed with a 1,048,576-token context length. Catalog observation only — DeepSeek has not been checked for a corresponding announcement, and the ID could be a snapshot alias or a listing that is withdrawn again.
The ID x-ai/grok-4.6 showed up in OpenRouter's public model-list API on 2026-08-21, listed at 500,000 context and named "SpaceXAI: Grok 4.6". This is a catalog observation only — no xAI announcement was checked or is claimed, and the entry could be a test, alias, or placeholder.
The model ID google/gemini-3.7-flash appeared in OpenRouter's public model-list API on 2026-08-21 with a 1,048,576-token context length. Catalog observation only — no Google DeepMind announcement was checked or is claimed, and the listing could still be withdrawn.
Codex CLI 0.149.0 adds an interactive codex agents dashboard for managing tasks, a codex queue command for messaging existing local or remote sessions, and TUI commands for changing the working directory. It also expands codex doctor diagnostics and hardens several MCP, auth-token, and managed-config paths.
Beta adds /handoff, moving a live local session — conversation, images, optional follow-up command — to a Cloud workspace that keeps running after the app closes, with a preflight that checks the repo, branch and commit are pushed. Cloud now lives in the existing Local/Remote environment menu and the separate Local/Cloud toggle is gone. The release also rolls up 0.0.14 stable, including the app rename to "Cline".
A wide maintenance release: plugin marketplaces and MCP configs can now mint HTTP headers via a headersHelper command (gated behind folder trust and stripped of inherited credential env vars), self-hosted runners get graceful-shutdown and proxy-authorization options, and unbounded memory growth in long interactive sessions is fixed. Most of the rest is Remote Control and cross-session messaging reliability. Anyone relying on the Ctrl+L double-press /clear shortcut loses it.
Mistral announced Agentic Search, described as a retrieval layer that lets AI systems navigate, read, and verify information inside complex documents. The source is a one-line announcement blurb with no API, pricing, or availability detail.
Oh My Pi 17.4.0 removes the global token-counting functions in @oh-my-pi/pi-agent-core and replaces them with immutable per-model Tokenizer instances, which several context-management functions now require explicitly. On the agent side it adds /cleanse, omp ps, speculative compaction, and an extendedContext setting controlling whether premium long-context windows are used.
Adds a built-in Concise output style that leads with results and skips preamble and narration while doing the work as thoroughly. Fixes prompt caching for sessions routed through an LLM gateway or custom base URL.
Preview release with notable infrastructure fixes: Cloud Workstations OAuth proxy redirect now resolves dynamically, capacity errors use context-aware silent retries with a TTL, cancelled multi-turn requests roll back fully, the TUI no longer hangs indefinitely (execution timeout added), and a directory mismatch in IDE connections is fixed. Also includes evaluation tooling (tool call formatter, failure summaries) and many SSR-agent issues (privacy notice wording, /clear context reset docs, fixed admin error for personal accounts, strict-null errors in tests, subagent output formatted as markdown).
Adds ANTHROPIC_DEFAULT_MODEL to set the starting model for new sessions while /model overrides persist across restarts. Adds notify_when_idle to SendMessage for one-shot cross-session coordination. Sandbox wildcard read-deny rules now correctly take precedence inside allowed regions. Auto mode gains severity-scored classification on all supported backends and a hardened git status check. VS Code gets screen reader support for the transcript. The /model picker trims to fit the terminal height. SIGTERM in print/SDK mode exits cleanly. Dozens of bug fixes across fullscreen rendering, session recovery, SendMessage, and more.
27 recorded relationships. A pulse travels only down an edge the archive holds — an entity with none ripples alone.
Significance 4+ across every vertical, sent 07:00 UTC. No summaries of summaries. 1,103 entries archived and counting.