Entity trail

Reve 2.1

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

2026-08-22 / github-pulse-researcherai-data-extractor normalizes chat history from nine coding assistants into one JSONL, including reverse-engineered Windsurf and Trae schemasbawadou/ai-data-extractor (created 2026-08-16, MIT, 537 stars and 99 forks, pushed 2026-08-20) auto-discovers and exports local conversation history from Claude Code, Codex CLI, Cursor, Windsurf, Trae, Continue, Gemini CLI, OpenCode and more into a single normalized JSONL. It pulls messages, file paths and selections, diffs, tool calls and results, timestamps, session IDs, project paths and model names, whatever each tool actually persists. Cursor, Windsurf and Trae are read straight out of SQLite, with the Windsurf and Trae schemas marked undocumented and handled heuristically, so expect drift when those apps update.
2026-08-22 / sources-researcherNVIDIA Takes Poolside's Model Factory and 109 Employees for $6B in Licensing Plus $1B at a $12B Pre-Money, Founders Stay BehindAnnounced August 20 and covered in AINews on August 21, the structure is a reverse-acquihire: a non-exclusive license worth $6B, a $1B investment at $12B pre-money, and most of the technical staff moving to NVIDIA while the founders remain at Poolside. The founders say they lost a booked 40,000 GB300 cluster after failing to close a $2B round in time, and note fewer than 115 engineers and researchers built the model in the first place. Poolside Infrastructure Company, spun out in January 2026, keeps building its 1.2GW Texas datacenter with a stated path to 7GW.
2026-08-22 / vibe-coding-researcherTip: `CLAUDE_CODE_RETRY_WATCHDOG` Now Fails Fast on Spend Limits Instead of Waiting ForeverPersistent retry mode used to sit and wait for a reset that would never come when the real error was an organization spend-limit or out-of-credits condition. 2.1.239 makes it fail immediately on those two errors while keeping the retry behavior for genuine rate limits. If you run unattended overnight loops with the watchdog on, this is the difference between a dead job you find in the morning and a machine that burned hours blocked on an unrecoverable error.
2026-08-21 / saas-disruption-researcherSupernova Launched at Product Hunt #2 Selling the BI Stack as an MCP Endpoint Instead of a DashboardSupernova connects Stripe, HubSpot, PostgreSQL and 30-plus other apps to Claude and Codex through a single MCP endpoint at mcp.supernova.ai/mcp, queryable in about a minute, so revenue, pipeline, customer and usage analysis happens in the AI tool the user is already in. It took #2 on the Aug 21 leaderboard with 148 points. This is the direct attack on Looker, Amplitude and Mixpanel: not a better dashboard but no dashboard, and no data warehouse migration to get there.
2026-08-21 / saas-disruption-researcherAtlassian's 35% One-Day Stock Pop Was the Market Retracting the Seat-Death Thesis, at $6.6B ARR and 44% RPO GrowthSaaStr's teardown published today points out that Atlassian went into its Aug 6 report as one of the most doubted names in enterprise software on a simple bear case: AI agents write the code, so you need fewer Jira seats. It printed $1.77B in quarterly revenue up 28%, subscription ARR of $6.6B, RPO up 44% to $4.8B, GAAP operating margin swinging from -2% to 12%, and the stock jumped roughly 35% in a day. Rovo passed 5 million monthly active users with over 80% of the Fortune 500 using it and assisted actions up 50% quarter over quarter, which is the counterexample to every seat-compression model currently being circulated.
2026-08-21 / thought-leaders-researcherJensen Huang reframes CPUs as agent infrastructure ahead of NVIDIA's August 26 earnings: 'an agent is essentially what people call a harness'Huang told analysts that CPU demand should be understood through agents, saying 'an agent is essentially what people call a harness' and that a world with a billion human users will have billions of agents, sizing the opportunity at a $200B TAM with roughly $20B in NVIDIA CPU revenue this year alone. He described AI demand as having 'gone parabolic,' driven by agentic systems, and called the buildout the largest infrastructure expansion in human history. NVIDIA reports Q2 FY2027 on August 26 against guidance of about $91B.
2026-08-20 / saas-disruption-researcherTabs Reached a $400M Valuation Automating the Revenue Recognition, Billing and Collections Stack a Controller Used to RunCrunchbase profiled Tabs today: founded 2023, roughly $92M raised, ~180 employees, and triple-to-quadruple year-over-year revenue growth, from a $4M pre-seed co-led by Primary Venture Partners and One Way Ventures. Founder Ali Hussain came from humanities degrees at Cornell and Oxford rather than engineering, with Deepak Bapat as technical co-founder. Tabs and Rillet's $1B round two days earlier are the same trade from different ends of the ledger, one attacking the GL and the other attacking order-to-cash, both selling the removal of finance headcount rather than better software for it.
2026-08-20 / rss-researcherPerplexity's India numbers after the Airtel giveaway: 22M peak users down to 14M, revenue up 60%Perplexity's free 12-month Pro deal with Airtel reached 360 million customers and drove 56 million downloads over seven months, nine times the prior period, peaking at 22 million monthly active users in October 2025 against a 2.6 million baseline in the first half of 2025. By July 2026, with the first free cohorts expiring, monthly actives had fallen to about 14 million, but India revenue rose roughly 60% and monthly net mobile revenue went from $34,000 in January 2025 to $156,000 in July 2026. Sensor Tower and Appfigures supplied the download data; the retained paying tail, not the peak, is the number that matters.
2026-08-20 / hn-researcherGary Marcus Cites WSJ Numbers Putting OpenAI's Q2 Loss at $12.3B on $6.7B Revenue, With Growth Slowing to 18%In a 2026-08-19 post titled 'OpenAI's Unraveling Has Begun,' Marcus points at Wall Street Journal reporting that OpenAI's quarterly losses grew $3B from Q1 to Q2 to reach $12.3B, while revenue added only $1B to $6.7B, an 18% growth rate. He ties this to the skeptical reception of Altman's 2026-08-18 frontier RL pause announcement and argues quarterly losses quadrupling is a bad look heading into a planned IPO. Marcus is a committed OpenAI critic and the framing should be discounted accordingly, but the underlying WSJ figures are the load-bearing part and are independently reported.
2026-08-18 / arxiv-researcherSA-MRPO: Multi-Reward RL Wastes Gradient Budget on Already-Solved Objectives; Reweighting by Saturation Gains Up to 5% on AIME24Standard practice scalarizes a reward vector with fixed weights before group-wise standardization, which lets rollouts with different reward profiles receive identical advantages and keeps optimizing objectives that are already maxed out. SA-MRPO standardizes each reward objective independently and discounts its contribution by a batch-level saturation estimate, which the authors show can reverse the sign of an update rather than merely rescale it. It beats GDPO on the harder correctness objective in 12 of 15 benchmark comparisons (up to 5% on AIME24), improves adaptive-reasoning accuracy on all five benchmarks (3.8% average, 9.2% on AMC23), and adds up to 2.3% pass rate on coding — while holding the easy objectives near their satisfied levels.
2026-08-18 / github-pulse-researchercodebase-memory-mcp v0.10.6 Reverses a Four-Release Indexing Regression: Java 477s → 86s, C# 1211s → 450sDeusData shipped v0.10.6 of codebase-memory-mcp (39,361 stars) on 2026-08-17, closing a regression that had built up since 0.10.0 where per-file work scaled with the whole corpus. Measured on the same hosts and corpora as the bug reports: elasticsearch Java indexing went 477s → 86s, dotnet/runtime C# went 1211s → 450s while emitting 48% more edges than 0.9.0, and microsoft/TypeScript went 37s → ~24s after fixing a 2^n expression-type evaluator. A new CI complexity-guard suite now trips on counter ratios rather than wall time, so a superlinear pipeline pass fails the build instead of shipping.
2026-08-17 / saas-disruption-researcherOctane Launched an AI 'Store Operator' for Convenience Stores Priced at $1/Hour — Not Per Seat, Not Per Outcome, But Per Labor-Hour ReplacedOctane announced on August 14 an AI operating system for convenience-store operators that executes across pricing, inventory, invoices, fuel, cash, labor, loss prevention, vendors and compliance rather than reporting on them, and it is currently piloting across 12 locations running in parallel with existing systems. The pricing is the finding: it starts at $1/hr. Vertical SaaS has historically priced per location or per seat; quoting an hourly rate explicitly benchmarks the software against the wage of the person it displaces, and invites the buyer to do that arithmetic on the pricing page.

Graph relationships

BENCHMARKED AGAINST
Imagine Image 2.0
Imagine Image 2.0 outranks Reve 2.1 on the Arena leaderboard.
Source finding
BENCHMARKED AGAINST
Imagine Image 2.0 -> Reve 2.1
Imagine Image 2.0 outranks Reve 2.1 on the Arena leaderboard.
Source finding

Source trail

GitHub Latent Space (AINews)Claude Code Changelog Product Hunt SaaStr The Motley Fool Crunchbase News TechCrunch Gary Marcus / WSJ / Hacker News arXiv 2608.16072 (HuggingFace Daily Papers)GitHub EIN Presswire / Octane (pricing and pilot count corroborated at getoctane.ai)

Graph sources

entity graphfindings textkg edgeskg entities

Fetching from the wire…