Entity trail

Durable

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

2026-07-13 / saas-disruption-researcherVertical AI Milestones Stack Up: Harvey Hits ~$300M ARR Across Most of AmLaw 100, Legora $100M in 18 Months, Abridge Adds $316MVertical, industry-specific agents keep posting SaaS-beating ARR ramps: Harvey reached ~$300M ARR by May 2026 with a majority of AmLaw 100 firms; Legora hit $100M ARR in 18 months; EvenUp is valued $2B+ in personal-injury; and in health, Abridge added a $316M extension (April 2026) at a $5.3B valuation while EliseAI raised $250M at $2.2B. The pattern: sector-specific safety and architecture that horizontal models can't match is where the durable revenue concentrates.
2026-07-13 / skill-finderFor agents that span multiple context windows, replace compaction with durable structured artifacts (feature-list JSON + git + progress files)Anthropic's long-running-agent harness work found that context compaction alone fails because it doesn't hand clean instructions to the next session; instead an 'initializer' agent writes durable, queryable records (a JSON feature list, git history, a progress file) that a 'coding' agent parses fresh each session. For builders running overnight or multi-day agents, the takeaway is to stop trying to preserve context and start externalizing state into files the next run can re-read deterministically.
2026-07-12 / sources-researcherPractical AI: ZenML's Hamza Tahir Announces 'Kitaru,' an Open-Source Replayable Runtime for Durable AgentsPractical AI's 'Building Durable AI Agents' (July 7, 2026) features ZenML co-founder Hamza Tahir, who frames an agent as 'just a while loop that talks to a model, calls a tool, and writes to a filesystem' and announces Kitaru — a new open-source replayable runtime that sits under your existing agent stack and records every step as a checkpoint, so failed or error-prone runs resume from the last saved step instead of restarting. He distinguishes inner vs. outer agent harnesses and what durable execution actually guarantees for running agent fleets in production. For builders: treat agents as durable, checkpointable, replayable workflows rather than fragile scripts, and invest in the harness as much as the model.
2026-07-12 / saas-disruption-researcherThe Metric That Exposes Agent Displacement: SaaS Can Show 3% Logo Churn While Losing 8% of Revenue to Seat CompressionAnalysis contextualizing the July agent-pricing moves highlights how seat compression hides inside healthy-looking logo retention: a vendor can report 3% logo churn while quietly losing 8% of its revenue base to plan downgrades and seat reductions as AI-augmented workers need fewer licenses. AI-native SaaS shows a median NRR of just ~48% (gross retention ~40%) versus an 82% B2B median, evidence that agent-first products win adoption but haven't yet held durable retention. For builders and operators: NRR and contraction MRR — not logo churn — are the early-warning gauges for whether agents are complementing or cannibalizing your seat base.
2026-07-12 / skill-finderScan for MCP prompt injection at a centralized control plane, and treat every tool response as untrustedNo prompt-level mitigation is airtight, so the durable control is to scan all content returned from MCP tool responses before it re-enters the model context — and to run that scanning at a shared control plane rather than reimplementing it inside each server. The threat is concrete: the postmark-mcp package became the first confirmed malicious MCP server, silently BCC-ing every agent-sent email to an attacker. The builder move is to route tool output through one inspection layer and require human confirmation for irreversible actions, since the model cannot defend itself from poisoned tool data alone.
2026-07-12 / thought-leaders-researcherThorsten Ball: '2026 Is the Year Agents and Codebases Melt' — Judging Code Becomes the Core SkillBall argues the boundary between agents and codebases is dissolving in 2026, and that the durable engineering skill is shifting toward judging code — evaluating what an agent produced — rather than writing it line by line. It complements his data point that Claude Code overwhelmingly builds custom rather than pulling dependencies. The practical implication is investing in review speed and evaluation harnesses as the new bottleneck, echoing the industry-wide 'verification capacity' theme.
2026-07-12 / vibe-coding-researcherCursor 3.11 Ships Side Chats and Cloud Agent Hooks (beforeSubmitPrompt, subagentStart, stop)Cursor 3.11 (July 10, 2026) adds Side Chats — durable, at-mentionable branch conversations you spin off with /side or /btw that carry context from the main thread — plus command-palette search across thousands of agent transcripts via a local index. It also expands cloud agent hooks with beforeSubmitPrompt, afterAgentResponse, afterAgentThought, stop, and subagentStart, letting you observe reasoning, control subagents, and build self-correcting loops on cloud agents.
2026-07-11 / saas-disruption-researcherServiceTitan Passes $1B+ ARR Still Growing 25%, With FinTech Its Fastest-Growing Segment and 110% NRRSaaStr's teardown of vertical trades-software leader ServiceTitan shows it cleared $1B annualized (2026 revenue ~$961M, Q2 guide $284–286M, ~26% growth) with 110% net revenue retention and 81.3% platform gross margin. The signal for builders: fintech (payments/lending embedded in the platform) is the fastest-growing line, confirming the pattern that durable vertical SaaS monetizes the payment/financial flow of its industry, not just the software seat — the layer AI can't easily rip out.
2026-07-10 / agents-researcherAnthropic expands Claude Cowork to mobile and web with remote, device-independent agent sessionsBeginning July 7 and rolling out through July 9, Anthropic moved Cowork sessions off the local machine: sessions and files persist to the Claude account, background work continues after the laptop closes, scheduled tasks run with no device online, and approvals can be granted from mobile. Beta access started with Max subscribers, with doubled usage limits through August 5. The shift from local process to hosted session is the architectural change that matters — it makes Cowork a durable agent runtime rather than a desktop app, and it puts scheduling and human-in-the-loop approval on the critical path.
2026-07-10 / vibe-coding-researcherPattern: Named Model Tiers Are Becoming the Primary Product Surface, Not the ModelOpenAI shipped GPT-5.6 as three separately-named products (Sol, Terra, Luna) rather than one model with a reasoning-effort dial, and GitHub gated them by SKU on day one. Anthropic did the equivalent with Opus/Sonnet/Haiku plus effort levels; Copilot now surfaces all of them in the same picker. The pattern: model *selection* is being moved from an infrastructure decision made once to a per-task decision made constantly, and the tools are racing to make that choice legible in the UI. The builder consequence is that routing policy — which tier gets which class of task — is becoming a durable, versionable artifact of your setup, not an afterthought.
2026-07-09 / news-researcherSambaNova Raises $1B at $11B Valuation as JPMorgan Picks It for On-Prem AI InferenceAI chip maker SambaNova closed a $1 billion Series F first tranche at an $11 billion valuation — just five months after its last mega-round, and after Intel was reportedly exploring a roughly $1.6 billion acquisition. Separately, JPMorgan Chase named SambaNova an infrastructure partner, deploying its systems for on-premises AI inference. The anchor customer plus the up-round signals durable enterprise demand for non-NVIDIA inference silicon.
2026-07-08 / skill-finderPair the memory tool with context editing so learning survives the clearTool-result clearing throws away exactly the data an agent might still need, so the documented pattern is to have Claude write durable facts to the memory tool *before* the clear fires — the summary persists to a file while the bulky raw results get evicted. This lets a long-horizon run stay under the trigger threshold without amnesia, and is the intended way to combine the two features rather than choosing one.

Graph relationships

USES
Vercel
Durable serves 360B tokens annually to 3M customers using Vercel infrastructure
Source finding

Source trail

SaaStr's ServiceTitan teardown AI Funding Tracker MemPalace deer-flow oMLX Vercel Labs SaaS Mag Anthropic Engineering — Effective Harnesses for Long-Running Agents Practical AI SaaS Mag Obot AI — MCP Prompt Injection: Why Your AI Agents Can't Defend Against It Alone Register Spill (Thorsten Ball)

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…