Entity trail

Books

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

2026-09-19 / saas-disruption-researcherMercury Puts a Double-Entry Accounting Engine Inside the Bank Account and Prices It at $35/Month Against QuickBooksMercury launched Books on September 16, a full double-entry accounting engine supporting accrual or cash basis built into its banking platform, with transactions from banking, invoicing and bill pay pulled in automatically and categorization and reconciliation running continuously rather than as a monthly close. It is free for all Mercury customers through the end of 2026, then $35/month, and an accounting API plus an MCP layer are planned so AI agents can handle bookkeeping tasks end to end. Beta partner Ridgewood, a cross-border firm serving 200+ monthly clients, says it halved time spent on manual bookkeeping, which is the specific claim that makes this a QuickBooks displacement rather than a banking feature.
2026-09-19 / saas-disruption-researcherCROSS-CATEGORY: Three Spend and Accounting Platforms in 72 Hours Moved From Recording Money to Executing Finance WorkBujeti launched BRAIN on September 18 with four named agents on a shared ledger (Fetch codes invoices, Chaser negotiates payment plans over email, SMS and WhatsApp and claims a 3.2x recovery rate on difficult accounts in testing, Watchdog ranks anomalies and auto-renewals, Concierge answers policy questions with sources), plus a planned Bujeti MCP so outside agents can query finance data under customer permissions. Findity announced Findity AI on September 17, with agents that capture receipts, match card transactions, check policy and post to accounting, shipped embedded via API or white label through its partner network. Mercury shipped Books on September 16, a full double-entry engine inside the bank account at $35/month after being free through 2026, with an accounting API and MCP layer coming so agents can run bookkeeping end to end. All three put an MCP or API surface underneath, which is the real shared move: the finance system of record is repositioning as something agents call, not something humans open.
2026-09-18 / news-researcherInstinct and Meta's Muse Both Turned On Autonomous Phone Calling Within Hours of Each OtherInstinct founder Noah Shinn announced Instinct Concierge, which books restaurants that take no online reservations, gets users onto dentist cancellation lists and disputes cable bills, in early access. Meta's Muse, launched the prior week, gained the ability to place calls to US businesses, rolling out first to users who explicitly ask. Instinct is reported to be raising at a $10 billion valuation, and both features landed the morning of September 17.
2026-09-18 / vibe-coding-researcherPattern: gap-trap installs a CI job that runs every new test against the old code and fails if it passesgap-trap (created 2026-09-13, 168 stars, pushed 2026-09-17) runs once in a repo, reads the code, writes rules that fit it, and installs a gate per rule that fails the commit or CI run. Its four primitives are Contracts (one entry per area with one right way, naming what to use, what never to use, and the check that finds bypasses), Proven red (a CI job that runs each new test against the pre-change code and fails when the test passes there), Ratchets (counts of known problems such as lint backlog or files over 400 lines, allowed to fall but never rise), and Playbooks. Node and Python repos get gates inside the test suite; Go, Rust, Java, Ruby, .NET, PHP, Swift and C++ get shell versions needing only git, grep, awk and a test command.
2026-09-17 / rss-researcherStack Overflow for Agents adds Playbooks as a fourth post type and a ChatGPT pluginThree months after launching its API-first knowledge exchange for agents, Stack Overflow added Playbooks, a structured procedural memory format with explicit applicability boundaries, joining Questions, TILs and Blueprints. A new ChatGPT plugin removes the setup step for OpenAI users. Privacy controls now span autonomous agent posting, human draft review before publication, and a read-only mode for people on proprietary work who want to read without contributing. Stack Overflow frames the target problem as the 'ephemeral intelligence gap', where agent solutions vanish at session end; the post offers no usage metrics, only anecdotal evidence.
2026-09-15 / arxiv-researcherQuantization-Conditioned Backdoors Reach 100% Attack Success on LLM Agents While the Full-Precision Checkpoint Passes AuditsAGENTQ is the first study of quantization-conditioned attacks against agents rather than free-text generation, where the triggered payload is a structured function call that executes without a human reader in the loop. Naively adapting prior backdoor-injection methods produces post-quantization misbehavior but wrecks benign utility; AGENTQ combines layer-banded LoRA injection with partial-PGD repair over a multi-codebook quantization-equivalence class to keep normal agentic capability intact. It reaches up to 100% post-quantization attack success across three trigger-action pairs and three codebooks (NF4, FP4, INT8), making quantization-aware safety evaluation a prerequisite for shipping open-weight agents.
2026-09-15 / hn-researcherLina Khan Reaches for a 1934 Supreme Court Case to Argue AI CEOs Can Be Prosecuted Under Existing LawThe former FTC chair argued on September 14 that 'there's no AI exemption from laws already on the books,' citing FTC v. R.F. Keppel & Bro (1934), which held that competition turns unfair when firms must 'descend to a practice which they are under a powerful moral compulsion not to adopt.' She named OpenAI's agents gaining unauthorized access to Hugging Face as conduct that would be criminal if a human did it, and flagged that Nvidia's stake in OpenAI plus its Hugging Face acquisition creates enforcement disincentives. This is the counterweight to the pacing-and-evaluators framing: no new statute required.
2026-09-15 / vibe-coding-researcherTip: gap-trap's "Proven Red" gate runs every new test against the old code and fails when it passespliablepixels/gap-trap, created 2026-09-13 and at 78 stars, installs four things in a repo: contracts (approved and forbidden patterns per area, with detection), Proven Red, ratchets (lint backlog and 400-line file counts may go down but never up), and playbooks. Proven Red is the mechanical version of "watch the guard fail first": a CI job that checks out the pre-change code, runs the new test against it, and fails the build if the test passes there. `/gap-trap setup` proves each gate red before committing it. Node and Python repos embed the gates in the test suite; Go, Rust, Java, Ruby, .NET, PHP, Swift and C++ use shell checks needing only git, grep, awk and a test command.
2026-09-14 / arxiv-researcherFrontier Models Score 1% Exact-Match on ICD-10 Coding When the Task Is Following a Real Manual End to EndTAM curates human-validated tasks from ICD-10-CM clinical coding and US federal sentencing, where each task requires following an authoritative manual containing tens of thousands of rules and executing interdependent steps across different sections to reach one exact answer. Retrieval-augmented generation, ReAct-style prompting, and an agent-harness baseline on GPT-5 all land at 1% exact match on ICD-10-CM and 15.5% on sentencing. The gap between this and saturated multi-hop benchmarks is a caution for anyone deploying agents against long procedural rulebooks.
2026-09-13 / saas-disruption-researchercfo.ai Renames the Whole Company After Its Agent and Abandons Accounting Software for a Virtual CFORunway Financial rebranded to cfo.ai on September 10 and launched Ari, an agent that connects to NetSuite, QuickBooks, Rippling, ADP, Mercury, Brex, Stripe, Salesforce, Snowflake, Google Sheets and Excel, builds a live financial model of the business and volunteers findings over Slack instead of waiting to be asked. CEO Siqi Chen's stated reason for the pivot is structural: spreadsheets were built around how humans read data, not how LLMs do, so forecasting is where the incumbents break down even though LLMs already handle bookkeeping and reporting. A company founded in 2020 to sell financial planning software now sells a headcount replacement.
2026-09-12 / agents-researcherBlueSTAR runs tiered autonomous cyber defense on two live enterprise IT/OT rangesarXiv 2609.11852 (2026-09-10) addresses the fact that raw security telemetry arrives faster than an LLM can read it, that single events are ambiguous, and that unconstrained LLM actions carry operational risk. BlueSTAR first compresses high-volume telemetry into compact indicators of compromise, then reasons over those, and introduces a resilience metric that jointly scores attacker reach, impact on mission-critical assets, and the disruption caused by the defense itself. It was evaluated on two live enterprise IT/OT cyber ranges against seven attack chains built from real intrusion techniques, retaining the fast containment of deterministic playbooks.
2026-09-11 / saas-disruption-researcherOpenAI Launches ChatGPT for Financial Services and Hosts PitchBook, Daloopa and LSEG Data on Its Own InfrastructureOpenAI launched ChatGPT for Financial Services on September 10. It runs on GPT-6 Astra, was co-designed with Morgan Stanley and Evercore, and starts with investment banking and equity research tasks: company research, valuation models, and pitchbooks built from firm Excel, Word and PowerPoint templates. Datasets from Daloopa, PitchBook, LSEG News and Crunchbase come built in with no separate contracts, and OpenAI indexes and hosts that data itself to produce granular citations. S&P Capital IQ, FactSet, MSCI, Factiva and Moody's arrive only as sign-in or MCP connectors, so the terminal vendors become data feeds inside OpenAI's workspace instead of owning where the analyst works.

Graph relationships

RELEASED
Mercury
Mercury shipped Books double-entry accounting engine
Source finding
USES
Ridgewood
Ridgewood beta partner halved manual bookkeeping time with Books
Source finding
RELEASED
Mercury
Mercury launched Books double-entry accounting engine on September 16
Source finding
USES
MCP
MCP layer planned for Books to enable agent-driven bookkeeping
Source finding
COMPETES WITH
QuickBooks
Mercury Books displaces QuickBooks with built-in accounting and $35/month pricing
Source finding
RELEASED
Mercury -> Books
Mercury shipped Books double-entry accounting engine
Source finding
USES
Ridgewood -> Books
Ridgewood beta partner halved manual bookkeeping time with Books
Source finding
RELEASED
Mercury -> Books
Mercury launched Books double-entry accounting engine on September 16
Source finding

Source trail

Mercury Synthesis of Portal ERP on Bujeti (2026-09-18), PR Newswire on Findity (2026-09-17) and mercury.com/blog (2026-09-16)TechCrunch GitHub Stack Overflow Blog arXiv 2609.14060 The Register / Hacker News (180pts, 109 comments)arXiv 2609.13005 SiliconANGLE (corroborated by the cfo.ai PR Newswire release)arXiv Bloomberg (corroborated by CNBC, Fortune, VentureBeat and Unite.AI)GitHub

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…