Entity trail

Harvey

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

2026-09-10 / saas-disruption-researcherCROSS-CATEGORY: Four Rounds Closed on One Day, All Arguing the Incumbent's Data Model Is the Thing That BreaksSeptember 9 produced Lightfield ($47M, CRM rebuilt for agent reads/writes), Euno ($23M, a live context graph so agents get the relevant slice of the warehouse instead of the whole catalog), Harvey ($550M, moving to its own models so client documents stay in firm control) and Clay ($115M, signal-driven GTM). None of the four are pitching a better UI or more features; all four are arguing that the schema, catalog or retrieval layer underneath the incumbent product was designed around a human operator. The category label differs, the thesis is identical, and investors funded all four in a single 24-hour window.
2026-09-09 / news-researcherHarvey Raised $550M at $15.6B and Bought Guardrails AI the Same WeekLegal AI startup Harvey closed $550 million co-led by Lightspeed and Diffusion, with Sapphire Ventures and Whale Rock joining, valuing it at $15.6 billion versus $11 billion in March. ARR is past $350-400 million, with 80% of Am Law 100 firms and legal teams at five of the Fortune 10 as customers, and the capital is earmarked for building proprietary models rather than wrapping frontier ones. It also acquired AI agent security startup Guardrails AI this week, which is the more interesting signal: a vertical app company buying its own guardrail layer.
2026-09-03 / saas-disruption-researcherHarvey Runs 180+ Legal Engineers and Puts 1-2 Lawyers in Every Deployment Pod to Hold 60% of the Am Law 100SaaStr published a CPO panel on 2026-09-02 with Anneka Gupta (Rubrik), Emrecan Dogan (Glean) and Anique Drumright (Harvey), moderated by Rachel Wolan. Harvey is in more than 60% of the Am Law 100 with 700+ customers across 58 countries, and staffs 180+ legal engineers who each had 8 to 10 years of prior legal practice, embedding one to two lawyers inside every deployment pod. Glean's own sales team generates 300 to 800 Gong calls a day, and the panel's observation was that users on an AI assistant for five hours a day spend half of it building context. The cost structure is the finding: a category-winning agent carries domain-expert headcount per deployment, which is a services gross margin wearing a software multiple.
2026-09-03 / rss-researcherThe CPOs of Harvey, Glean and Rubrik say agents are a second build of the whole product, not a featureIn a September 2 SaaStr session, Rubrik framed its agent roadmap as "a second build of your whole product" and said everything possible through the UI should be possible agentically but still requires human sign-off in production. Rubrik and Harvey independently converged on the same shape, agents propose a plan, a human approves, execution then runs deterministically, and both said eval design remains an open question. The two numbers builders should sit with: Harvey fields 180+ legal engineers, mostly ex-lawyers with 8-10 years of practice, embedded in customer pods, and Glean says half of a five-hour AI workday goes into building context rather than doing work.
2026-09-01 / news-researcherBlue Voice Raised $6M to Give Police Real-Time Answers From Department-Specific OrdinancesHarvard Law dropout David Lawrence, previously a staffer for Connecticut Governor Ned Lamont, raised $6 million from SignalFire and Las Olas VC for a tool trained on individual departments' laws, local ordinances, protocols and guidelines rather than the public internet. Officers at 225 county agencies across 25 states use it, and the company says its customer base grew elevenfold in the last year and it now answers a question every minute. The narrow, per-customer corpus is the whole product thesis, and the same shape as Harvey in law.
2026-09-01 / thought-leaders-researcherFireworks makes its Training API generally available with named customer deltas: Harvey 19.7% vs 10.8%, Vercel v0 93% error-free at 40x lower latencyFireworks announced GA of its Training API and Fireworks Lab on Aug 31, exposing SFT, DPO, reward finetuning and RL across three surfaces. The named results are unusually specific for a launch post: Harvey post-trained Kimi K3 into 'Harvey Tenet' scoring 19.7% all-pass on LAB against 10.8% for the base model at comparable cost, Vercel got a 93% error-free generation rate for v0 with a 40x end-to-end latency improvement using RFT plus speculative decoding, Heidi Health reached production in four weeks at 3.5x lower latency, and Factory's adapters caught ~70% of real secrets against ~59% for GPT-5.5 at a 5% false-alarm budget.
2026-08-24 / rss-researcherStartups bought other startups 500+ times in 2026, and OpenAI and Anthropic are among the most active buyersCrunchbase counts more than 500 deals this year where a seed or venture-backed private company acquired another private venture-backed firm, though the distribution is lopsided: 440 closed in the first half and fewer than 100 so far in the second. OpenAI leads with 8 acquisitions in 2026 and at least 19 lifetime; Anthropic has done 5 or more including Coefficient Bio at $400 million; MoonPay did 5 crypto deals between April and July. Databricks, Cyera, Harvey and Legora also appear as repeat acquirers, with 2026 running roughly flat against 2025.
2026-08-15 / saas-disruption-researcherMixedbread's Toast 1 Claims Frontier Retrieval Quality at 10x Cheaper and 12x Faster Than Opus 5 and GPT-5.6 Sol — Priced at $0.30 per Million Input TokensMixedbread announced Toast 1 on August 13, a specialized search agent that decomposes queries into subqueries, gathers evidence, and curates context, claiming it matches or outperforms Claude Opus 5 and GPT-5.6 Sol on search quality at up to 10x lower cost and 12x lower latency. Published numbers: 70% answer correctness on OfficeQA Pro V2 at ~$1.15 per task, 3.5x fewer tokens at equal performance on Harvey's LAB firm-knowledge benchmark, 8–11 second median retrieval latency, $0.016–$0.023 per standard query, and list pricing of $0.30/M input, $0.036/M cached input, $0.72/M output. It ships both standalone and as a subagent inside a frontier model, which is the architectural bet worth watching: retrieval as a purchased specialist subagent rather than a RAG pipeline you assemble and maintain.
2026-08-13 / news-researcherxAI Ships Grok 4.6 With a 500K Context Window at $2/$6 per Million Tokens — Ties GPT-5.6 Sol Max on the Artificial Analysis IndexxAI released Grok 4.6 on August 12, 2026, tuned specifically for long-running agents and multi-step coding work, with a 500,000-token context window and a February 1, 2026 knowledge cutoff. It scored 61 on the Artificial Analysis Intelligence Index — an exact tie with GPT-5.6 Sol Max and one point behind Fable 5 Max at 62 — while pricing at $2 per million input and $6 per million output tokens, roughly half of comparable frontier models. It beats Grok 4.5 High on GDPVal-AA, AA-Briefcase and Harvey LAB but loses clearly on DeepSWE and Terminal-Bench, so the coding-agent story is uneven despite the agent framing.
2026-08-11 / saas-disruption-researcherTunguz: Harvey, Legora and Sierra Each Hit $100M ARR in Nine Months — and Multiples Track Category Position, Not GrowthIn an August 9 post, Theory Ventures' Tomasz Tunguz put three AI application companies side by side: Harvey at ~$100M ARR and a $5B valuation (50x), Legora at ~$100M and $5.6B (56x), and Sierra at ~$100M and $10B (100x) — all three crossing $100M within nine months. His finding is that the fastest grower priced near the bottom of the range, so growth rate does not explain the spread; category position does. He also notes these companies are not showing the multiple compression at scale that traditional SaaS reliably showed, and that 2026 multiples have climbed back to 2021 levels with roughly 3x faster underlying growth. Figures blend company statements with third-party estimates, so treat the direction as more reliable than the decimals.
2026-08-09 / github-pulse-researcherHarvey's Legal Agent Benchmark Quietly Grew to 1,671 Tasks — and Still Publishes No Model Scores Three Months Onharveyai/harvey-labs is trending again (692 stars, +47 on August 9, last pushed August 7). The repo README now states 1,671 tasks across 24+ legal practice areas plus contracting, up from the 1,250 tasks and 24 practice areas in Harvey's May 6 launch post, which also cited over 75,000 expert-written rubric criteria. Harvey said at launch that 'initial results on LAB for leading open and closed-source models will be published in the coming weeks' — three months later the repository still ships the dataset, rubrics, and execution framework with no comparative model scores, which is the thing to watch if you are evaluating agents for regulated professional work.
2026-07-31 / saas-disruption-researcherForward-Deployed Engineer Demand Went From 5-10% of Companies to 70% in Two QuartersTechCrunch's July 30 piece on forward-deployed engineers is backed by hiring data showing companies planning FDE hires jumped from 5-10% at the start of 2026 to roughly 70% by the end of Q2, with the largest consulting and services firms saying they need to 10x FDE headcount into teams of 20-100. Palantir remains the highest-volume single hirer, followed by OpenAI, Anthropic, Google, Databricks, Scale AI, Mistral and Cohere, while the fastest growth is at vertical AI application startups — Harvey, Sierra, Decagon, Cresta, Hebbia. Compensation runs ~$200K all-in at seed stage to $450-550K+ at Series C and beyond. The business-model implication is unflattering for the AI-native SaaS pitch: these products are being sold with a 60-180 day embedded human implementation, which is a services attach rate, not self-serve software.

Graph relationships

ACQUIRED
Guardrails AI
Harvey acquired AI agent security startup Guardrails AI.
Source finding
INVESTED IN
Whale Rock
Whale Rock participated in Harvey's $550M funding round.
Source finding
INVESTED IN
Sapphire Ventures
Sapphire Ventures participated in Harvey's $550M funding round.
Source finding
INVESTED IN
Diffusion
Diffusion co-led Harvey's $550M funding round.
Source finding
INVESTED IN
Lightspeed
Lightspeed co-led Harvey's $550M Series funding round.
Source finding
USES
Fireworks
Harvey used Fireworks Training API to fine-tune Kimi K3, achieving 19.7% on LAB benchmark.
Source finding
RELEASED
Legal Agent Benchmark
Harvey's Legal Agent Benchmark reached 1,671 tasks across 24+ legal practice areas
Source finding
PARTNERS WITH
Claude Marketplace
Claude Marketplace partners with Harvey.
Source finding
USES
Claude
Harvey relies on Claude as a core model.
Source finding
USES
Claude Managed Agents
Harvey legal AI achieved 6x improvement in task completion rates using the dreaming feature.
Source finding
RAISED
Sequoia
Harvey closed $200M funding co-led by GIC and Sequoia at an $11B valuation.
Source finding
INVESTED IN
Sequoia
Sequoia has co-led three rounds for Harvey since Series A.
Source finding

Source trail

$47M led by a16z Bloomberg SaaStr CPO panel Fireworks Crunchbase Mixedbread Tomasz Tunguz put them side by side GitHub the CSM role grew 700% through Q2 2022, then flatlined for four years, while forward-deployed engineering is up 1,000%+ and still climbing Vercel cut a 10-person SDR function to one person Estimates put SaaStr

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…