Developers on the Frontline of the SaaS Replacement Wave  🔗 **Source:** [Developers on the Frontline of the SaaS Replacement Wave](https://leaddev.com/technical-direction/how-developers-navigate-the-c-suites-war-on-saas) > [!note]+ Document Notes > tags: [saas-replacement, ai-coding-tools, build-vs-buy, data-sovereignty, developer-tooling, software-economics] > author: [MISSING] > source: LeadDev > reference: https://leaddev.com/technical-direction/how-developers-navigate-the-c-suites-war-on-saas > date created: {{date:YYYY-MM-DD}} {{time:HH:mm}} > > # 🏷️ Title: Developers on the frontline of the SaaS replacement wave (How developers navigate the C-suite's war on SaaS) > <!-- deep-read:sections start --> ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ### 🔗 References - [SaaS is dead, long live Service-as-a-Service (Forbes)](https://www.forbes.com/sites/alexanderpuutio/2025/07/04/saas-is-dead-long-live-service-as-a-service/) - [Klarna shuts down Salesforce (Seeking Alpha)](https://seekingalpha.com/news/4144652-klarna-shuts-down-salesforce-as-service-provider-workday-to-meet-same-fate-amid-ai-initiatives) - [6AM City's secret weapon: 400 newsletters with no staff](https://www.amediaoperator.com/news/6am-citys-secret-weapon-400-newsletters-with-no-staff/) - [Zach Lloyd / Warp LinkedIn post on pausing SaaS](https://www.linkedin.com/posts/zachlloyd_weve-stopped-buying-saas-and-and-are-moving-activity-7443305541709422593-Abx4/) - [Retool AI build vs buy report 2026](https://retool.com/blog/ai-build-vs-buy-report-2026) - [Sebastian Siemiatkowski on the 20VC podcast](https://www.youtube.com/watch?v=P7vIRAFSXmk) ## 📝 Summary - **What happened.** Reporters at LeadDev document a real but uneven "SaaS replacement" wave: C-suites are re-scrutinizing their SaaS stacks and, with AI-coding assistants, replacing some tools with in-house or self-hosted builds. Klarna is the flagship case, having shut down Workday and Salesforce; Warp, 6AM City, and a Berlin startup (DmarcDkim.com) show more measured variants. - **Why it matters.** The economics of "buy a subscription" have shifted. When a usable replacement can be built in ~2 days, the bar to justify recurring SaaS spend rises sharply. This reshapes vendor relationships, procurement, and what developers spend their time on. - **Key stakeholders.** Engineering leaders and CTOs (cost + sovereignty), developers (now builders/maintainers of replacements), SaaS vendors (facing churn and "artificial moat" scrutiny), and security/compliance teams (inheriting new risk). - **Risks and tradeoffs.** Data extraction from closed-source SaaS is the real bottleneck, not building the replica. Self-hosting open-source clones can introduce unvetted vulnerabilities. Maintenance shifts from vendor to internal teams. Some AI initiatives *add* SaaS rather than remove it. - **Bottom line.** SaaS isn't dead — but the justification bar is much higher, and the hard part of leaving is your data, not your code. ## 📖 Main Content The core argument: AI-coding tools plus geopolitical/data-sovereignty pressure have made in-house replacement of SaaS economically viable for a growing subset of tools, but the wave is selective and carries underappreciated risk. - **The trigger conditions.** Two forces combine: capable AI-coding assistants (lowering build cost) and rising concern over cost, data sovereignty, privacy, and governance — especially in Europe. Individual employees empowered to experiment with AI drive a lot of the internal-tool replacement bottom-up, not just top-down mandates. - **What's actually easy vs hard.** Building a replica and even productionizing it is not the hard part. The hard part is *data transfer* — SaaS tools are closed-source with internal databases, so proprietary data is "stuck in the SaaS provider's data model." Klarna's CEO frames switching cost of data as the decisive barrier; once solved, it becomes a real threat to SaaS. - **The "artificial moat" claim.** Some vendors charge extra for interacting with or extracting data. As companies adopt agentic solutions that need to touch that data, they hit friction — putting those very tools most at risk of reevaluation. (Contested framing: "artificial" is the interviewee's characterization; a vendor would call it product/pricing design.) - **Security is the underestimated cost.** Grabbing "open source" projects off GitHub and deploying them can introduce unvetted vulnerabilities, missing security certifications, or vulnerable middleware. Mitigations cited: choose open-source with active codebases (Antypov), and build a shared deployment+security layer so per-app security work is rare (Warp/Desai). - **Maintenance nuance.** Antypov distinguishes infrastructure maintenance (~1 day/month) from software maintenance, and *deliberately chooses not to maintain the software itself* — because that's where the economics stop working. - **Not universal.** Vibe coding doesn't always remove SaaS; one of Lampe's clients onboarded *more* SaaS because AI simplified integrating logging tools. The honest conclusion: the calculus now favors internal tools more often, but SaaS survives where product benefits justify it. Terms: **SaaS** = Software-as-a-Service (subscription-hosted software). **Agentic development** = building with AI agents that act, not just autocomplete. **Data sovereignty** = legal/physical control over where data lives and who can access it. **Vibe coding** = building software largely by prompting AI rather than hand-writing code. Uncertainties: the survey figures come from Retool (a low-code vendor with an interest in the build narrative); Klarna and Retool both declined/didn't respond to interviews, so the flagship examples rest on secondary reporting. ### 🗒️ Notes - Klarna (late 2024) shut down several SaaS providers including Workday and Salesforce for a lighter stack; declined interview. - 6AM City estimates saving **$100,000/month** replacing its CRM with an AI-powered in-house version. - Warp paused SaaS purchases; CEO Zach Lloyd estimates saving **>$10,000/year** in canceled subscriptions. - Warp moved its marketing site off Framer in-house for A/B flexibility beyond Framer's API limits (Aloke Desai, head of product engineering). - Warp hackathon produced a docs replacement for GitBook with brand styling, telemetry hookup, and auto-generated docs from releases — "maybe like two days of work." - Retool survey: **35% of 817 customers** already replaced one SaaS tool with a custom build; **78%** expect to build more of their own tools in 2026. Retool didn't respond to interview. - DmarcDkim.com (Berlin, founder Oleksii Antypov) replaced ~**10 SaaS** with self-hosted OSS: Rocket.chat over Slack/Teams, Twenty over HubSpot/Salesforce. - Felix Godbout (infrastructure architecture consultant): C-suite pushing on-prem for data sovereignty; enterprises eliminate SaaS via AI automation or cloud functions — but risk of unvetted vulnerabilities. - Godbout on data transfer breakage: forms deep inside ERP, automated processes in "system XYZ" all need recreating; risk of duplicating data across two platforms during migration. - Jess Lampe (global lead technologist, Launch Consulting): agile-era lifecycle tools not always suited to agentic development; "artificial moats" via data-access charges; but one client onboarded MORE SaaS via AI. ## 📚 References - [SaaS is dead, long live Service-as-a-Service — Forbes](https://www.forbes.com/sites/alexanderpuutio/2025/07/04/saas-is-dead-long-live-service-as-a-service/) - [Klarna shuts down Salesforce — Seeking Alpha](https://seekingalpha.com/news/4144652-klarna-shuts-down-salesforce-as-service-provider-workday-to-meet-same-fate-amid-ai-initiatives) - [6AM City's secret weapon — A Media Operator](https://www.amediaoperator.com/news/6am-citys-secret-weapon-400-newsletters-with-no-staff/) - [Zach Lloyd on pausing SaaS — LinkedIn](https://www.linkedin.com/posts/zachlloyd_weve-stopped-buying-saas-and-and-are-moving-activity-7443305541709422593-Abx4/) - [Retool AI build vs buy report 2026](https://retool.com/blog/ai-build-vs-buy-report-2026) - [Sebastian Siemiatkowski on 20VC — YouTube](https://www.youtube.com/watch?v=P7vIRAFSXmk) ## 🗣️ Paraphrase - Thought leaders predicted SaaS's death when AI-coding tools arrived; a few years on, C-suites are actually reevaluating their stacks under cost, geopolitics, and AI pressure. - Klarna is the loudest example (dropping Workday, Salesforce); others like Warp, 6AM City, and DmarcDkim.com are moving more cautiously, citing five- and six-figure savings. - The genuinely hard part isn't building or shipping the replacement — it's extracting proprietary data locked in closed-source SaaS databases. - Security is the underrated risk: casually deploying unvetted "open source" can import vulnerabilities; smart adopters pick active OSS and centralize security. - SaaS isn't dead, but the bar to justify a subscription is now much higher, and the answer tilts toward internal tools where clear product benefits exist. ## 📰 Newsletter The "SaaS is dead" headlines are back, but this LeadDev piece is refreshingly grownup about it: yes, teams like Klarna and Warp are ripping out subscriptions and saving real money, but the hard part of leaving a tool isn't rebuilding it — it's prying your own data out of a closed-source database. For the WordPress crowd, there's a familiar lesson here: open source only wins if you actually vet it (deploying a random GitHub repo can quietly import vulnerabilities), and "self-hosted" means *you* now own maintenance and security. Worth a read if you're weighing build-vs-buy for internal tools in 2026. Links: the [Retool build-vs-buy 2026 report](https://retool.com/blog/ai-build-vs-buy-report-2026) and [Klarna's CEO on switching costs](https://www.youtube.com/watch?v=P7vIRAFSXmk). ## ✅ Action Items #### DevRel - Developer advocacy: publish a build-vs-buy decision guide for internal tools, with the "data extraction is the hard part" caveat front and center. - Education: run a workshop on safely vetting OSS before self-hosting (certifications, active codebase, middleware audit). - Developer experience: template a "shared deployment + security layer" pattern so replacement apps inherit security instead of rebuilding it. - Developer marketing: content series on real savings numbers vs hidden maintenance costs. - Community management: survey your community on which SaaS tools they've replaced and what broke. - Internal/external/cross-project: align procurement and engineering on a SaaS-justification bar. #### OSPO - Policy & Governance: define a policy gate for deploying third-party OSS internally (no unvetted GitHub deploys). - License Compliance: scan self-hosted replacements for license and middleware obligations. - Security Management: stand up a vuln-intake path for internally hosted OSS clones. - Software Inventory: SBOM every self-hosted replacement; track freshness of upstream OSS. - Developer Support: office hours on migrating data out of closed-source SaaS. - Community Engagement: fund/upstream the OSS projects you now depend on. - Strategic Alignment: report SaaS-replacement savings AND absorbed maintenance/security cost to execs. #### Community - Maintainers: publish hardening guides for the OSS tools being adopted as SaaS replacements. - Contributors: "first issue" rotations aimed at data-export/import tooling. - Sponsors: transparent impact reports linking corporate self-hosting to upstream funding. - Ecosystem: coordinated disclosure drills for widely self-hosted replacements. #### WordPress - Core: N/A to core directly, but frame WordPress self-hosting as a mature model for the "own your stack" trend. - Core Performance: profile self-hosted alternatives against SaaS baselines. - Core AI: responsible-AI guardrails for AI-generated internal tools. - **Core Program Team:** track build-vs-buy narrative's effect on ecosystem plugins. - Mobile: N/A — not applicable to this source; article is stack/economics-focused. - Playground: reproduce "replace a SaaS in two days" as an education demo. - Community: meetup talk on data-sovereignty-driven self-hosting. - Training: lesson path on vetting and hardening OSS before deploy. - Marketing: narrative tying WordPress ethos to the SaaS-replacement moment. - Support: knowledge-base entry on migration data-loss pitfalls. - Design: N/A — not applicable to this source; no design surface discussed. - Documentation: examples-first guide to auto-generating docs from releases (Warp's pattern). - Photos/TV/Polyglots: N/A — not applicable to this source; no media/localization angle. - Accessibility: ensure self-built replacement UIs meet WCAG 2.2 AA (a gap when speed-building). - Meta/CLI/Hosting/Tide/Openverse/Themes/Plugins: Hosting — cross-host checklist for self-hosting OSS replacements safely. #### The WPCC - Programs: microgrants for maintainers of OSS tools now used as SaaS replacements. - Projects: documentation sprint on secure self-hosting. - Funding experiments: matched grants tied to corporate self-hosting adoption. ### FAIR - Technical independence: data-portability standards so leaving SaaS doesn't hinge on the vendor's data model. - Policies and covenants: anti-lock-in covenants (no "artificial moat" data-extraction fees). - Productized governance artifacts: a self-hosting security scorecard. ## 🧪 Hypotheses - H1: Teams that replace SaaS with self-hosted OSS underestimate total cost of ownership once security and data-migration labor are counted. - H2: The primary predictor of a successful SaaS exit is data portability, not the difficulty of rebuilding features. - H3: Vendors that charge for data extraction ("moats") will see higher churn in the agentic era than vendors with open data access. - H4: A meaningful fraction of "AI replaces SaaS" initiatives net-increase SaaS usage (integration/logging) rather than decrease it. - H5: Two-day AI-built replacements accrue latent maintenance/security debt that surfaces 6–18 months later. ### 📊 Methodology - H1/H5: longitudinal case studies tracking labor hours and incident counts pre/post replacement over 12–18 months. - H2: regression across migrations with data-portability score as predictor of success/time-to-switch. - H3: cohort analysis of churn segmented by vendor data-access pricing. - H4: audit net SaaS line-items before/after AI initiatives across a sample of firms. ### ❓ Questions - What percentage of "SaaS replacements" survive 24 months without reverting? - How much of the reported savings is offset by internal salary/maintenance cost? - Do self-hosted OSS replacements meet the same compliance certifications the SaaS held? - Is the Retool survey (vendor-run) representative, or selection-biased toward builders? ### 🔮 Future Research - Independent (non-vendor) measurement of build-vs-buy outcomes and reversion rates. - Data-portability standards and tooling as the true unlock for SaaS competition. - Security posture comparison: managed SaaS vs self-hosted OSS clones at scale. ### 💡 Implications - **Practice:** engineering leaders should gate self-hosting behind vetting + a shared security layer, and budget for absorbed maintenance. - **Policy:** data-portability and anti-lock-in becomes a procurement and possibly regulatory lever. - **Security:** self-hosting shifts a large, often-invisible risk in-house. - **Education:** developers need OSS-vetting and data-migration skills, not just build skills. - **Funding models:** corporate self-hosting raises the stakes for sustainably funding upstream OSS. ### 🧠 Learning Styles Assessment - Reading level estimate: Flesch-Kincaid Grade ~11–13 (business-journalism register, multi-clause sentences, some jargon). Flesch Reading Ease roughly 40–50. #### 👀 Visual learners - Present: two figures (author illustration, event promo) — no explanatory diagrams. - Improve: a build-vs-buy decision tree, a "what breaks during data migration" flow, before/after cost tables. #### 🔊 Aural learners - Present: references a 20VC podcast episode (external audio). - Improve: a 2–5 minute audio summary of the savings-vs-risk tradeoff. #### 🗣️ Verbal learners - Present: clear quotes and named practitioners; terms mostly defined in context. - Improve: a short glossary (agentic, data sovereignty, vibe coding, artificial moat). #### 🧩 Kinesthetic learners - Present: none — no hands-on steps. - Improve: a checklist for safely self-hosting an OSS replacement. #### 🧠 Logical learners - Present: implicit cost/risk tradeoff structure; survey stats. - Improve: a decision matrix (portability × maintenance × security) with scoring. #### 🧑‍🤝‍🧑 Social learners - Present: multiple practitioner voices. - Improve: discussion prompts for team build-vs-buy reviews. #### 🧘‍♀️ Solitary learners - Present: reflective "raise the bar" conclusion. - Improve: self-assessment on which of your SaaS tools clears the new bar. #### 🌿 Naturalistic learners - Present: none. - Improve: lifecycle framing (data as an ecosystem you migrate, not just copy). ### 📊 Accessibility Assessment #### Screen reader semantics - Present: figure/figcaption used; alt text exists on images (e.g., "LDX3 New York is live"). - Improve: the author-illustration alt text is a filename ("Karis_Harrington_developeranalyticsbuyersguide2024") — replace with descriptive alt. #### Keyboard navigation - Present: standard article links, presumably reachable. - Improve: verify visible focus and skip-link on the live page (not determinable from content). #### Contrast and readability - Present: not determinable from HTML content. - Improve: confirm 4.5:1 body text contrast and adequate line-height on the live site. #### Text resizing and reflow - Present: responsive image `sizes`/`srcset` present. - Improve: verify no clipping at 200% zoom. #### Media and captions - Present: figcaption on the event image. - Improve: the linked podcast should have a transcript for non-audio users. #### Forms and interactions - Present: N/A — no forms in article body (a register CTA exists). - Improve: ensure the registration gate is accessible and labeled. #### Standards reference - Gaps map to WCAG 2.2 AA 1.1.1 (non-text content — filename-as-alt) and 1.4.3 (contrast, to verify). ### 🌍 DEIB Assessment #### Cultural breadth - Present: multi-region examples (US Klarna/Warp/6AM City, Berlin DmarcDkim.com, EU sovereignty framing). - Improve: voices from Asia, Africa, Latin America; non-Western regulatory contexts. #### Inclusive language - Present: gender-neutral, jargon largely defined. - Improve: plain-language callout for "agentic" and "vibe coding." #### Neurodiversity support - Present: short paragraphs, key-takeaways bullets up top, consistent structure. - Improve: more subheadings to chunk the middle of the article. #### Representation - Present: several named practitioners across company sizes and roles. - Improve: gender/geographic balance among quoted sources appears skewed male; broaden. #### Psychological safety - Present: neutral, non-inflammatory framing; acknowledges counterexamples. - Improve: none needed materially. ### ✅ Logical Validity #### Strong reasoning - Causal (well-supported): lower build cost (AI) + data-sovereignty pressure → higher SaaS-replacement rate. The article grounds this in multiple independent cases, strengthening the inductive claim. - Inductive generalization with a stated limit: uses Klarna/Warp/6AM City/DmarcDkim.com to infer a trend, then explicitly qualifies it ("not universal"), which is epistemically honest. - Distinction (strong): Antypov's infra-maintenance vs software-maintenance split is a clean conceptual clarification that supports the cost argument. #### Fallacies - Hasty generalization risk: the headline "war on SaaS" over-extends from a handful of vocal cases; mitigated (not eliminated) by the "not universal" caveat. Fix: lead with base rates, not flagship anecdotes. - Loaded language / question-begging: "artificial moats" presumes the pricing is illegitimate rather than a design choice. Fix: attribute clearly as one interviewee's view and present the vendor rationale. - Appeal to (interested) authority / survey bias: leaning on Retool's vendor-run survey (35%/78%) to size the trend. Fix: cite independent data or flag the conflict of interest inline. #### Tone of voice - Measured trade-press analysis: "It's not dead, but the bar is much higher" typifies the balanced, hedged tone. ### 🛡️ DARVO Analysis - Deny: N/A — not applicable to this source. It's third-party journalism, not a defendant's rebuttal; no denial pattern present. - Attack: N/A — not applicable to this source. No attacks on accusers or credibility-undermining; the "artificial moats" line critiques a practice, not a person. - Reverse victim/offender: N/A — not applicable to this source. No attempt to recast an accused party as the victim. ## 🎯 Conclusion - Bottom line: SaaS isn't dead, but AI-coding tools plus data-sovereignty pressure have raised the bar to justify a subscription — and the genuinely hard, underestimated parts of leaving are data extraction and security, not building the replacement. - Practical next steps: treat data portability as the first question in any build-vs-buy call; vet OSS before self-hosting and centralize security; budget honestly for maintenance you're now absorbing; and discount vendor-run "everyone's building" surveys. ## 🔄 Refactor Appendix — Developers on the frontline of the SaaS replacement wave > Density: 2–5 Literature atoms (H3), 0–3 Permanent candidates (H4). ### 📚 Literature Atoms (H3) #### [H1] Data extraction, not rebuilding, is the SaaS exit bottleneck > [!quote] ID:rw_manual1 > It's not building a replica or even the productionization that is challenging. Often it's the data transfer because these SaaS tools are closed-source **Context & use**: - Mid-article, Godbout (infrastructure architecture consultant). - Reframes build-vs-buy: the moat is the data model, not the feature set. **Tags**: #xref **Highlight**: [view highlight](readwise://) <!-- atomicity: 1 idea ✓ --> #### [H2] AI raised the bar to justify a SaaS subscription > [!quote] ID:rw_manual2 > It's not dead, but the bar is much higher to justify a SaaS subscription. **Context & use**: - Concluding argument of the piece. - The one-line thesis for any procurement conversation. **Tags**: #xref #aphorisms **Highlight**: [view highlight](readwise://) <!-- atomicity: 1 idea ✓ --> #### [H3] Casual OSS self-hosting imports unvetted vulnerabilities > [!quote] ID:rw_manual3 > it just outright introduces vulnerabilities within their own environment because it's not vetted properly **Context & use**: - Godbout on the security cost of "it's open source, let's deploy it." - Security is the underestimated line item in replacement economics. **Tags**: #xref **Highlight**: [view highlight](readwise://) <!-- atomicity: 1 idea ✓ --> #### [H4] Choose to maintain infrastructure, not the software > [!quote] ID:rw_manual4 > We are deliberately choosing not to maintain the software itself, because that's where the price point wouldn't make sense anymore. **Context & use**: - Antypov distinguishes infra maintenance (~1 day/month) from software maintenance. - The cost model that makes self-hosting viable. **Tags**: #xref **Highlight**: [view highlight](readwise://) <!-- atomicity: 1 idea ✓ --> ### 🧠 Permanent Candidates (H4) #### [P1] Data portability is the real competitive moat in the agentic era **Claim** (one sentence, Courtney's voice): Whether a company can leave a SaaS vendor depends on how portable its data is, not on how hard the software is to rebuild. **Evidence pointer**: [[#H1]] **Scope & conditions**: Holds for data-heavy, workflow-embedded SaaS (CRM, ERP); breaks for stateless or thin-integration tools where switching cost is low regardless of data. **Links**: - [[Build-vs-buy decisions]] — because extends - [[Vendor lock-in]] — because applies-in procurement - [[Open standards for data export]] — because evidence-for <!-- atomicity: 1 idea ✓ --> ### ✏️ Fleeting Queue - [F1] Track reversion rate: how many SaaS replacements survive 24 months? — no independent data in article #fleeting #inbox - [F2] Is the Retool 35%/78% survey selection-biased toward builders? — vendor-run #fleeting #inbox ### → Things 3 (max 2 actions) - [ ] Draft a build-vs-buy decision guide leading with data-portability — content #things - [ ] Add an OSS self-hosting security checklist to the knowledge base — ospo #things <!-- deep-read:sections end --> 📄 **Full text:** [[Full Document Contents/Articles/Developers on the Frontline of the SaaS Replacement Wave.md|Developers on the Frontline of the SaaS Replacement Wave by leaddev.com]] ## Highlights - 35% of its 817 customers have already replaced one SaaS tool with a custom build ([View Highlight](https://read.readwise.io/read/01kx436seh8b3j5zt8vrza63j5)) - Note: Key trend stat (note: vendor-run Retool survey — sourcing caveat). - Tags: atom/literature/h5 - it just outright introduces vulnerabilities within their own environment because it’s not vetted properly ([View Highlight](https://read.readwise.io/read/01kx436nkp6yax60rfrbd12vpp)) - Note: Security is the underestimated cost of casual OSS self-hosting. - Tags: atom/literature/h3 - We are deliberately choosing not to maintain the software itself, because that’s where the price point wouldn’t make sense anymore. ([View Highlight](https://read.readwise.io/read/01kx436xm3y48hz76ph8ze6e41)) - Note: Infra-vs-software maintenance split that makes self-hosting viable. - Tags: atom/literature/h4 - It’s not building a replica or even the productionization that is challenging. Often it’s the data transfer because these SaaS tools are closed-source ([View Highlight](https://read.readwise.io/read/01kx436ap99gg8t7n9jhkabk8w)) - Note: Reframes SaaS exit: data extraction is the bottleneck, not rebuilding. - Tags: atom/literature/h1 - It’s not dead, but the bar is much higher to justify a SaaS subscription. ([View Highlight](https://read.readwise.io/read/01kx436e9qwhbn750jk89yn24q)) - Note: One-line thesis of the piece. - Tags: atom/literature/h2