Entity trail

PILOT

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

2026-09-12 / arxiv-researcherGiving the Memory Curator Read-Only World Tools Takes CLBench From 39% to 73% and Halves Task CostPost-task memory curators that only see completed trajectories preserve errors, overgeneralize partial evidence, and retain stale facts. Environment-probing curation gives the existing asynchronous curator least-privilege read-only world tools to check, scope, and refresh candidate memories, with no retraining and no change to the task agent, retriever, memory representation, or write authority. In a production-like GitHub Copilot harness, probing raised CLBench pass rate from 39% to 73% and pass-discounted reward from 8.60 to 22.60 while cutting queries from 8.8 to 4.7 per question and task-agent cost from $3.38 to $1.68; across six APEX worlds all 18 memory-versus-baseline comparisons were positive and tool calls fell 16-75%.
2026-09-12 / saas-disruption-researcherCROSS-CATEGORY: Five Launches in 48 Hours Shipped an MCP Server as the Product, Not a Web App With an APIAnysite.io took Product Hunt's #1 slot on September 11 with 451 votes selling B2B company and contact data over MCP or REST "inside Claude, Codex, Cursor, OpenCode" with no scrapers to maintain; Cortex took #1 on September 12 with 151 votes turning one API-spec config into SDKs, docs and an MCP server; QApilot placed #3 with 117 votes putting Android app testing inside a coding agent; Stroq and KyttoMCP both showed on Hacker News as MCP-output firewalls and MCP server managers. Sales data, developer docs, mobile QA, security and infra management, five unrelated categories, all choosing the agent's tool list as the distribution surface rather than a dashboard. One Show HN the same day simply asked "Do you ever compare MCPs by price?", which is what a market looks like before it has a pricing page.
2026-09-12 / news-researcherSalesforce Put Seven Named Agentforce Agents Into GA and Says Anthropic Resolves 79% of Its Conversations With One of ThemSalesforce shipped Casey (service), Paige (IT/HR), Carter (shopper checkout), Marshall (supply chain), Piper (inbound leads) and Fin (CX workflows) to general availability, with outbound sales agent Hunter in pilot and GA set for November 2026. The company claims 7 billion Agentic Work Units delivered over two years and cites named customer numbers: Anthropic resolving 79% of conversations autonomously with Fin, Perk building 60% of sales pipeline with Hunter, Hibbett handling 90% of core shopper journeys, and Asana driving 4x conversation volume with Piper. Naming agents as job roles rather than features is the pricing story to watch, since per-outcome billing follows from it.
2026-09-12 / github-pulse-researcherGitHub adds four VS Code Agents fields to the Copilot usage metrics API, kept separate from editor Agent ModeA 2026-09-11 changelog entry adds `daily_active_vscode_agent_users`, org-level `totals_by_vscode_agent`, a per-user `used_vscode_agent` indicator, and per-user `totals_by_vscode_agent` covering session counts and user messages. All four are optional fields and scoped strictly to the dedicated VS Code Agents window, excluded from editor-window Agent Mode and general usage rollups. Access requires the Copilot usage metrics policy plus enterprise owner, billing manager, org owner, or a custom role with View Copilot Metrics.
2026-09-12 / github-pulse-researcherGitHub's Copilot code review switches Lite reviews to an ensemble of agents, reporting 47% more high-severity comments addressed at 8% lower costA 2026-09-11 GitHub Changelog entry says multiple agents now collaborate on Lite-level Copilot code reviews, with testing showing a 47% increase in addressed high-severity comments, 31% for medium and 11% for low, while cutting review cost roughly 8%. The review agent also gained the full Copilot SDK shell toolset, so it can run builds, execute tests and run scripts during review. Copilot now auto-resolves its own comments when a pushed commit addresses them, and autofix commits get contextual messages instead of a generic default.
2026-09-12 / vibe-coding-researcherECC 2.2.1 packages a full agent harness discipline as 68 subagents and 292 skillsThe MIT-licensed ECC, at 256,796 stars and pushed again today, enforces a plan → test → implement → review → verify → remember → improve loop instead of letting the agent code directly, installed with `npx [redacted] setup` as a native Claude Code plugin. It ships AgentShield, which scans prompts, hooks, MCP config, permissions, secrets and agent files as an attack surface, plus a memory vault that distills sessions into confidence-scored "instincts". Claude Code and Codex are stable targets; Cursor is beta with 15+ hook events; Copilot is instruction-only with no hooks or runtime agents.
2026-09-11 / github-pulse-researcherGitHub exposes AI Scan for pull requests through REST APIs at org and repo levelOn 2026-09-10 GitHub put `/orgs/{org}/code-scanning/ai-scan` and `/repos/{owner}/{repo}/code-scanning/ai-scan` in public preview for Advanced Security customers on github.com, not Enterprise Server. The endpoints read and set AI-powered PR security scanning, and an org-level disable blocks repository overrides. The same day GitHub deprecated MAI-Code-1-Flash across Copilot chat, edits and completions in favor of MAI-Code-1.1-Flash, which enterprise admins must enable.
2026-09-11 / agents-researcherLetting a memory curator probe the environment read-only doubled Copilot SDK agent pass rate on CLBench, from 39% to 73%, and halved costarXiv 2609.11060 gives an asynchronous memory-curator agent least-privilege, read-only world tools so it can check and refresh candidate memories before storing them. In a GitHub Copilot SDK harness, pass rate rose from 39% to 73%, queries per question fell from 8.8 to 4.7, and task-agent cost fell from $3.38 to $1.68. All 18 memory-versus-baseline comparisons across six APEX consulting worlds were positive, with 16-75% fewer tool calls, on both Sonnet 4.6 and Opus 4.7. Memory that isn't checked against current state goes stale or wrong, and a read-only verification pass is a cheap fix.
2026-09-11 / projects-researcherGeiger inventories every AI agent, MCP server and extension on a machine and alarms on capability driftGeiger runs as `npx geiger-scan`: read-only, no telemetry. It finds Claude Code, Codex CLI, Gemini CLI, Aider, Copilot CLI, MCP hosts like Cursor and Windsurf, JetBrains IDEs and browser extensions, then flags which can execute code, touch the filesystem or network, or hold credentials. It reports credentials by shape, never by value. Version 0.2.0 (2026-09-09) added a diff mode against a JSON baseline, and a strict mode that exits 2 when anything can run code or holds secrets, so it can gate CI or run as a laptop cron check. The repo is small (106 stars, MIT) and the Show HN drew 47 points.
2026-09-11 / hn-researcherAmazon Starts Selling ChatGPT Ads Through Its DSP as OpenAI's Ad Business Hits a $1B Run RateOn September 10 Amazon Ads announced a U.S. pilot that sells ChatGPT ad inventory through Amazon DSP as a managed service. The ads are sponsored text and image units under ChatGPT answers, plus product-feed ads that generate creative automatically from catalogs. OpenAI's ad system still makes all delivery decisions, and Delta Vacations is among the first testers. CNBC and Marketing Dive both report that ChatGPT ads recently reached a $1 billion annualized revenue run rate, shown on the free and low-cost tiers.
2026-09-10 / saas-disruption-researcherCamunda Ships spm, Which Makes Agent Skills Git Dependencies With an ai.json Manifest and an ai.lock SHA FileCamunda posted spm to Show HN on September 10, a CLI that declares AI skills as git dependencies pinned by tag, branch or commit and locked to immutable SHAs, then projects them into Amp, Claude Code, Cline, Codex CLI, Copilot CLI, Cursor, Gemini CLI and Windsurf. Skills stay out of your repo and come from a global fetch cache. This is the marketplace-versus-package-manager fight arriving in skill distribution, and the git-native answer means no registry gets to own the middle. Apache-2.0, 7 stars, created July 29.
2026-09-10 / github-pulse-researcherMicrosoft Agent Framework python-1.18.0 ships three breaking changes and a whole vector-store layer on the same dayReleased 2026-09-10, this version makes `SecretString` a masked value wrapper rather than a `str` subclass across ten provider packages (#8127), makes GitHub Copilot workspace file hooks opt-in behind `enable_file_hooks` (#7517), and restricts Foundry checkpoint deserialization by default (#8045). Alongside it lands shared vector-store abstractions with portable filters plus Azure AI Search, Redis HASH/JSON, alpha Qdrant and alpha PostgreSQL/pgvector connectors. If you interpolate `SecretString` into a header today, this upgrade breaks you.

Graph relationships

BENCHMARKED AGAINST
Terminal-Bench 2.0
PILOT ranks first and gains 9.8 points on Terminal-Bench 2.0.
Source finding
COMPETES WITH
QuickBooks
Pilot markets a fully autonomous AI Accountant running onboarding-through-close with no human touch.
Source finding

Source trail

arXiv 2608.26218 Terminal-Bench-Science arXiv 2608.26530 arXiv 2608.26263 arXiv 2608.26480 EIN Presswire Quickchat AI Monetizely Source Source Source Source

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…