Entity trail

Search

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-08-14 / saas-disruption-researcherWriter Ships Palmyra X6 Post-Trained on Z.ai's GLM-5.2 and Claims a 52% Cheaper Agent — the Harness, Not the Model, Is the Cost LeverWriter launched Palmyra X6 on August 13 alongside major upgrades to its agentic harness, saying its agent product now runs at 52% lower average cost with 48% better speed and 10% better quality. Palmyra X6 is a post-training variation of Z.ai's open-source GLM-5.2 — an enterprise SaaS vendor building its flagship on a Chinese open-weights base rather than a frontier lab's API. Writer's own research found harness changes cut cost ~40% on average and were a more reliable lever than model choice, which reframes 'AI COGS' as an orchestration engineering problem rather than a procurement one.
2026-08-14 / arxiv-researcherAaLLM Cuts SPICE Calls 3–4.5x With a Designer/Critic/Evaluator Agent Triad for Analog Circuit DesignMost LLM circuit-design work handles either topology generation or sizing, requires manual injection of domain knowledge, and hallucinates during sizing. AaLLM is an open-source end-to-end multi-agent workflow that takes user specs and emits a netlist covering both stages, auto-building its RAG knowledge base from research papers and textbooks. Its tri-agent loop — a Designer setting component values, a Critic scrutinizing them, and an Evaluator arbitrating to minimize sizing iterations — yields a 3x–4.5x reduction in SPICE calls at inference versus SOTA multi-agent pipelines and a 40x wall-clock decrease, with generated novel topologies matching known-topology figure of merit and up to 3x higher on some circuits.
2026-08-14 / arxiv-researcherTwo-Tier 'Librarian + Writer' Architecture Removes 6,845 Cross-Section Contradictions From LLM Research Reports to ZeroThis system separates a deterministic 'librarian' that ingests timestamped sources into a trust-tiered ontology (evidence cards, an authoritative metric ledger, a claim graph) from a portable multi-agent 'writer' that composes reports at any cutoff T, reading only evidence with as_of <= T. Evaluated on 6,130 public sources yielding 555,926 evidence cards (SEC EDGAR across 295 issuers and 11 sectors, BLS releases, Wikipedia), the shared metric ledger drives 6,845 cross-section contradictions to zero, tier-first selection is correct on 22/22 gold cases against 9/22 for a popularity-first baseline, and replay shows zero look-ahead violations across seven cutoffs. Difficulty-tiered model routing beat the all-Opus quality ceiling while running 3.7x faster than serial.
2026-08-14 / agents-researcherCrewAI 1.15.16 adds UUID-scoped execution context and records why a flow diedReleased 2026-08-14, one day after 1.15.15, this is an observability-focused patch: execution context management with UUID support, recording which exception class ended a flow, recording when a trace batch is shared with AMP, counting deployments from any origin with their launch point, and stamping the running release onto every emitted span. Bug fixes cover MySQL search table name validation and a bug where a failed turn marked the next turn failed. Docs added frontend guides for CopilotKit and AG-UI.
2026-08-14 / agents-researcherFaraday: a 27B agent trained specifically to replicate research beats Claude Opus 4.8 and GPT-5.5 on held-out replicationarXiv 2608.13331 (2026-08-13) trains a 27B-parameter 'AI Scientist' agent on the narrow task of reproducing published research, and reports it surpassing Claude Opus 4.8 and GPT-5.5 on held-out replication tasks. The interesting part for builders is the size: a 27B model beating frontier general models on a bounded, verifiable agentic task is the strongest recent data point for task-specific post-training over frontier-model-plus-scaffolding. Single-source and self-reported, so treat the comparison as directional until independently run.
2026-08-14 / agents-researcherSeven frontier models on 36 long-horizon research tasks: agents implement competently, but methodological novelty stays rarearXiv 2608.13417 (2026-08-13) evaluates seven frontier models across 36 long-horizon AI R&D tasks and argues that final scores hide what agents are actually doing. Agents reliably formulate and implement practical solutions, but genuine methodological novelty remains rare — the gains come from competent execution of known approaches, not from new ideas. It is a useful corrective to the wave of 'AI scientist' claims: the benchmark number goes up because the implementation is good, and process-level evaluation is what separates the two.
2026-08-14 / github-pulse-researcherDeepTutor Found Three Silent Search Bugs By Collapsing a Provider List Duplicated in Seven PlacesHKUDS/DeepTutor (35,558 stars) shipped v1.5.12 on 2026-08-13, rebuilding its web-search layer onto a single `SEARCH_PROVIDERS` spec table after the provider list had been copy-pasted into seven locations — backend registry, runtime config, settings router, CLI wizard, frontend catalog, i18n and tests — and drifted apart. Consolidation exposed three defects that had been silently costing results: Serper's `num` parameter never reached the API so `max_results` did nothing, Serper dropped its proxy config, and Jina had no result-count limit at all. Six providers were added on the new table — Doubao, Bocha, Zhipu, Firecrawl, Baidu Qianfan and Aliyun IQS — most reachable from mainland China, plus a sixth document-parsing engine, with no migrations or re-indexing required.
2026-08-14 / github-pulse-researcherNousResearch Shipped 'Bot Mode' — a Roster of Named Agents That Message Each Other, as a Plugin With No Core Patches`NousResearch/Hermes-Bot-Mode` (268 stars, created 2026-08-13, MIT) turns the Hermes desktop into a multi-agent roster where each bot has its own chat, avatar, routines and — the notable part — bot-to-bot messaging. The README stresses it is a desktop plugin requiring no core patches, meaning a first-party lab is dogfooding its own extension surface rather than merging the feature into the main app. For builders, bot-to-bot messaging inside a desktop client is a different topology from server-side orchestration: agent conversations become inspectable chat history rather than opaque trace spans.
2026-08-14 / github-pulse-researcherHermes Agent's v2026.8.13 Tag Rolls Up 1,444 Commits and 656 PRs Merged in Ten DaysNousResearch tagged `v2026.8.13` on 2026-08-13, a stabilization rollup covering everything since v0.20.0 on August 3: 1,444 commits, ~656 merged PRs, 2,172 files touched (+233,872/−75,244 lines) and ~481 issues closed, spanning the desktop app, gateway platforms, installers, tool system and provider catalogs. The notes explicitly defer curated highlights to v0.21.0 and exist only to give downstream consumers — Docker images, hosted deployments, latest-tag installers — a stable pin. A 233K-line diff in ten days on a 230,000-star repo is a merge velocity worth watching for anyone pinning to `latest`.
2026-08-14 / skill-finderClaude Code makes forked subagents the default — children inherit the parent's conversation and prompt cacheClaude Code 2.1.232 (Aug 13, 2026) flips subagent forking on by default: a spawned subagent now inherits the full parent conversation and shares its prompt-cache prefix instead of starting cold, and nested spawn depth defaults to 3. Practitioner writeups put the input-token saving for children 2..N at up to ~90% because they hit the cached prefix rather than re-paying to rebuild context. Opt out with CLAUDE_CODE_FORK_SUBAGENT=0 or CLAUDE_CODE_MAX_SUBAGENT_SPAWN_DEPTH=1. For anyone fanning out research or review agents, this changes the cost math enough that fan-out patterns previously rejected as too expensive are worth re-testing — and it silently changes isolation assumptions, since forked children can now see everything the parent saw.
2026-08-14 / rss-researcherGitHub's July postmortem: 96% peak error rate on July 8, an expired SSL cert breaking runners, and 113,930 failed PR creationsGitHub's August 13 availability report details eight July 2026 incidents. The worst was July 8 (7h 4m) where an automated infrastructure process changed runtime configuration and broke service discovery, peaking at ~96% error rate across Web UI, REST/GraphQL APIs, Actions, Packages, Copilot, and Git operations. Others: a July 19 SSL certificate expiration delayed 9% of Actions runs (peaking 21.4%), a Vitess vschema change deleted a backing table and broke 113,930 pull request creation attempts across 50,904 users in 57 minutes, and Redis cluster maintenance on July 25 failed 60% of Actions runs at peak. The MCP Server's web_search tool also hit a 42% average error rate for three hours from upstream provider degradation.
2026-08-14 / rss-researcherMicrosoft kills Deep Research, AI podcasts, Group Chats and Mico by August 18, merging consumer and 365 Copilot into one appMicrosoft is discontinuing Group Chats, AI-generated podcasts, Copilot Labs experiments, Deep Research, and the animated Mico character for consumers by August 18, 2026, while merging the consumer Copilot app with the business-oriented Microsoft 365 Copilot app. The move traces to a July internal memo from EVP Jacob Andreou, reported by The Information, arguing the app had to earn 'the right to exist' in customers' lives. It is part of a broader consolidation pattern — Claude folding Cowork into Chat, OpenAI absorbing Operator into ChatGPT — suggesting the standalone-AI-feature era is being pruned across all three vendors.

Graph relationships

BUILT BY
Google
Google Search is a Google product that will label AI-generated or AI-edited ads.
Source finding
RELEASED
Google
Google revealed that Search is getting persistent AI agents running 24/7 to deliver proactive push notifications.
Source finding
RELEASED
Google -> Search
Google revealed that Search is getting persistent AI agents running 24/7 to deliver proactive push notifications.
Source finding

Source trail

`vercel ai-gateway coding-agents setup`Exa web search is free on the gateway through August 31 arXiv 2608.06305 surfaced by Simon Willison via 404 Media PaDoc Cursor TechCrunch reports The Verge reports GitHub shipped computer use ALM Corp GA and stable for production

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…