Augment

1.1 Identity and business model

Secondary profiles consistently place Augment's founding in 2022 and describe an enterprise-focused AI coding platform for teams working in large, complex codebases. Gartner lists Palo Alto as the head-office location, while IntelPilot characterizes the model as B2B SaaS and reports a shift toward usage-based pricing. The product is distributed through developer workflows rather than a standalone greenfield-only proposition: reviewed sources describe IDE, command-line, code-review, and context services aimed at multi-repository engineering environments. These points establish the historical operating identity, but they do not establish the legal entity, product ownership, or office footprint after the reported October 2026 asset sale.[CO001, CO002, CO003, CO014, CO016]

1.2 Capital history and public scale signals

The clearest historical capital marker is the April 2024 emergence from stealth: multiple secondary sources report $252 million raised in total, including a $227 million Series B, and a $977 million valuation. Those figures are financing history, not evidence of current cash, revenue, or enterprise value. Public operating metrics are materially weaker. IntelPilot reports roughly 156 employees and $155.7 million of 2026 revenue, while Vantaige reports roughly $20 million of late-2025 ARR; the periods and metric definitions differ, and CheckThat says no public revenue or ARR/MRR figures exist. The apparent conflict requires primary financial evidence before either figure is used for valuation.[CO006, CO007, CO008, CO009, CO010]

1.3 Leadership, governance, and concentration

Vantaige and AI Central identify Igor Ostrovsky and Guy Gur-Ari as the technical founders and Scott Dietzen as a senior founding leader. Their backgrounds span large-scale storage architecture, Google AI research, and enterprise software leadership, which is directionally aligned with the product's large-codebase thesis. Current authority is not reliably established: one profile calls Dietzen executive chairman, another calls him chief executive, and a third records different CEO names while explicitly noting that the relationship is unclear. No assigned source supplies a current board roster, ownership table, voting rights, or post-acquisition management structure, leaving governance and key-person dependence as material diligence items.[CO004, CO005, CO016]

1.4 Milestones, product validation, and current continuity

Augment's public chronology shows rapid financing followed by agent and context product expansion: an agent launch in April 2025, two pricing-model changes during 2025, and Model Context Protocol access for the Context Engine in February 2026. Independent user-review evidence is positive on contextual suggestions but also flags irrelevant output, limited enterprise analytics, and the need for manual intervention. The latest corporate event is more consequential than another product release: Parsers.vc records Harness acquiring selected products, the Context Engine, and related technology, with the team joining Harness. Because the assigned pool lacks the underlying transaction announcement and terms, this chapter treats Augment's present standalone status, retained assets, liabilities, and capitalization as unresolved rather than carrying forward its pre-transaction stage.[CO011, CO012, CO013, CO014, CO015, CO016]

2.1 Market boundary and category position

The relevant market is not all generative AI or all developer productivity software. It is the spend on AI-enabled software-engineering workflows: code understanding, agentic implementation, code review, debugging, migration, onboarding, and the controls needed to deploy those workflows inside an organization. The inclusion boundary covers software subscriptions and usage budgets paid by engineering organizations for those activities. It excludes general-purpose chat, model infrastructure sold to developers, conventional IDE licenses, and unassisted consulting labor. Augment's public positioning is unusually specific within that boundary: the pool describes a platform built around multi-repository context, autonomous agents, and broader SDLC integration, while a 2026 SiliconANGLE report describes its Context Engine MCP as a way to make semantic code context available to other agents. This creates a plausible wedge between generic completion and enterprise codebase intelligence, but the evidence is largely product and market commentary rather than audited category revenue. The category should therefore be treated as an adoption opportunity, not as a verified company revenue pool.[CM001, CM002, CM003, CM004]

2.2 Sizing through multiple lenses

The evidence supports three different sizing lenses, none of which should be silently combined. First, a third-party market view in the assigned pool says the AI coding assistant market was projected at $20–30 billion by 2025, while another calls it a $6 billion market in 2026 growing at 22% CAGR. Neither fetched page exposes a methodology, and the incompatible vintages and values should not be normalized or treated as Augment's obtainable market. Second, a bottom-up price lens is observable: public plan descriptions put entry pricing at $20 per month, a team plan at $60 or $100 per month depending on the reviewed page and plan vintage, and enterprise pricing at custom rates. These are list-price anchors, not realized ARR or wallet share. Third, an evidence-constrained serviceable wedge is defined by teams with large or multi-repository codebases, complex migrations, regulated-data requirements, or a need for cross-tool context. The pool provides buyer and use-case descriptions for that wedge, but no verified count of eligible companies, seats, conversion, or net retention. Accordingly, the opportunity may support enterprise software economics, but the investable sizing remains a diligence gap.[CM005, CM006, CM007, CM008, CM016]

2.3 Buyers, adoption drivers, and constraints

The primary economic buyer is likely a VP of Engineering, CTO, or engineering platform leader; senior engineers, technical leads, architects, and security teams are users or influencers. The buyer's budget case is strongest when context loss, migration backlog, code-review throughput, onboarding time, or compliance controls are visible operating constraints. The reviewed buyer descriptions point to mid-sized and large software organizations with multiple repositories and legacy or polyglot systems, while Gartner peer material in the pool records positive user ratings but also mentions occasional irrelevant suggestions, limited analytics, and manual intervention to avoid hallucination. That combination matters: the adoption trigger is not simply interest in AI, but a measurable workflow bottleneck that justifies a new line item and human oversight. Switching costs can rise through indexed code, workflow integrations, memories, and review processes, yet those same integrations create data-governance and portability questions. Pricing uncertainty is material because usage-based inference fees, top-ups, and differences between public plan descriptions make annual budget predictability harder to assess. Finally, the October 2026 acquisition of selected Augment assets by Harness, as recorded by the assigned sources, is a market-structure event whose effect on existing plans and buyer confidence was not publicly clarified in those sources; it is a current diligence question rather than evidence of either acceleration or deterioration.[CM009, CM010, CM011, CM012]

2.4 Sizing and adoption gaps

Public material is sufficient to map the product category and likely buying motion, but not to underwrite market share. The missing denominators include the number of target engineering organizations, eligible seats per organization, paid versus trial conversion, realized price after discounts, usage expansion, renewal and multi-homing behavior, and the share of deployments that require enterprise controls. The pool also contains conflicting or time-sensitive commercial descriptions: one reviewed page lists Indie, Standard, Max, and Enterprise tiers, while another describes Standard and Business plans and a later asset-acquisition event. These differences may reflect plan changes or source quality, so they should not be normalized into a single price curve. The prudent market stance is therefore “credible category wedge, unverified serviceable market.” A buyer interview set, current price card, cohort-level retention data, and an independent market methodology would materially improve confidence.[CM013, CM014, CM015]

2.5 Exhibits

3.1 Competitive landscape and buyer alternatives

The relevant competitive set is broader than named AI coding vendors. Augment's direct peers include GitHub Copilot Enterprise, Cursor, Sourcegraph Cody, and Codeium/Windsurf; adjacent options include Tabnine, Continue.dev, Cline, Aider, and JetBrains AI Assistant. The status quo is conventional IDE tooling, internal developer-platform work, or a mixed stack in which an organization keeps its incumbent assistant and adds specialized context services. Gartner describes Augment as Developer AI that uses a large-codebase understanding to produce personalized suggestions and insights, while a 2026 alternatives comparison explicitly frames Cursor as the closest IDE-integrated alternative and Cody as the enterprise-scale codebase-context alternative. This means the buying decision is primarily about repository scale, workflow integration, security/deployment, and whether the buyer wants one agent or a composable stack. Augment's own comparison page is useful for identifying its intended category, but it is company-issued and not independent proof of superiority. More importantly, a low-reputation news index dated October 9, 2026 says Harness announced an acquisition of selected Augment assets, including Cosmos, Auggie CLI, and the Context Engine. That current signal could reshape the competitor map and product ownership, but the assigned pool lacks the primary transaction announcement or contractual scope needed to confirm whether Augment remains a standalone platform.[CP001, CP002, CP003, CP004]

3.2 Capability, distribution, and switching comparison

The central product trade-off is depth of context versus distribution and workflow breadth. A fetched comparison of Augment, Copilot, and Cursor says Augment's Context Engine can span 500,000 files and 30 or more repositories, while Copilot is described through a 64K-128K token window and Cursor through single-repository indexing; these are third-party review assertions and should be validated in a controlled pilot rather than treated as audited benchmarks. Another source describes Copilot's GitHub integration as its default advantage, Cursor's single-repository fit as its strength, and Cody's self-hosted option as a genuine differentiator. Gartner review evidence adds favorable user sentiment for contextual suggestions and IDE integration, but also records occasional irrelevant or generic outputs and limited enterprise analytics. Augment therefore has a credible wedge for complex, multi-repository teams, while incumbents retain distribution, ecosystem, and simpler onboarding. Switching costs are likely to rise with indexed repositories, memories, integrations, and workflow conventions, but the same integrations can support multi-homing; a team can pair Augment with another completion or IDE product.[CP005, CP006, CP007, CP008, CP009]

3.3 Pricing, packaging, and moat durability

Pricing is a material competitive variable and a source of uncertainty. The 2026 Fast.io review reports that Augment removed inline completions from non-Enterprise plans on March 31, 2026 and made Business the cheapest plan at $100 per month for up to 50 seats; it also reports unpredictable credit consumption and a 30,000-credit trial requiring a card. The same review contrasts this with free or lower-priced entry points for Copilot and Cursor. These details may be current but are one review's account and should be rechecked in procurement. A separate 2026 comparison lists Cursor at $20-$40 per user per month, Copilot Workspace at $10-$39, and Cody at custom enterprise pricing, while describing self-hosted options for Cody, Tabnine, and Codeium. Augment's durable advantage would need to come from measurable productivity, onboarding, review, or risk outcomes that outweigh price and lock-in; public evidence here is mostly company-selected testimonials, directory summaries, and reviewer assertions. The adverse case is that completions commoditize, pricing changes erode trust, and platform vendors bundle sufficient context with existing developer workflows.[CP010, CP011, CP012, CP013, CP014]

3.4 Competitive diligence verdict

Augment appears best positioned for engineering organizations with large, interdependent repositories, compliance requirements, and a willingness to pay for agentic workflow depth. Cursor is a credible single-repository and IDE-native substitute; Copilot is the strongest distribution and ecosystem incumbent; Cody and self-hosted alternatives are credible where deployment control dominates; open-source or BYO-model tools remain substitutes for teams prioritizing flexibility and auditability. The evidence does not yet establish a durable moat: independent win-loss data, realized pricing, retention, deployment breadth, and apples-to-apples productivity outcomes remain unavailable in this pool. The reported Harness transaction also makes product continuity, support ownership, and roadmap commitments immediate diligence items rather than ordinary future risks. A buyer should first verify the transaction scope, then run a parallel pilot on the same repositories, compare context-retrieval quality and total cost, test security/deployment requirements, and measure renewal intent before assuming Augment's positioning is defensible.[CP003, CP015, CP017]

3.5 Exhibits

4.1 Revenue model and pricing architecture

The clearest current public model is a hybrid subscription-and-consumption structure. A third-party June 2026 pricing review reports a Business plan at $100 per month for as many as 50 seats, with the same amount included for pooled usage; Enterprise is custom-priced. Usage economics are more explicit than realized revenue: model tokens are passed through at provider list prices with a 40% service fee, while Cosmos sandbox compute is reported at $0.19 per hour. This gives Augment three possible economic levers—base subscription, model-service markup, and compute—but public sources do not disclose customer-level discounts, overage incidence, Enterprise commitments, or revenue recognition. The bridge therefore maps billing mechanics rather than claiming a verified revenue mix or gross-profit conversion.[CI001, CI002, CI003, CI004]

4.2 Traction and sales-efficiency signals

Public traction is directional rather than underwriteable. GetLatka reports $20 million of annual revenue for October 2025 but explicitly says the source was not recorded; AI Central separately labels approximately $20 million as ARR in late 2025. Revenue and ARR are not interchangeable, and another retained directory says no public revenue or ARR data exist. Customer-review evidence is more useful for value discovery than for financial measurement: Gartner reviewers describe workflow benefits but also limited analytics and occasional irrelevant output, while one compiled customer example reports a shorter merge cycle. These signals can support willingness-to-pay hypotheses, but they do not reveal CAC, sales cycle, expansion, renewal, concentration, or payback.[CI006, CI007, CI008, CI012, CI013]

4.3 Cost structure and margin path

The disclosed billing design exposes important cost drivers without proving margin. Model-token cost is largely variable and visible to the buyer, the 40% service fee is intended to fund the Context Engine and platform, and Cosmos compute is billed separately. SiliconANGLE reports Augment's claim that better context can improve coding-agent quality and reduce wasted tokens, which is a plausible customer-value mechanism but not evidence of Augment's own gross margin. Hosting, inference rebates, support, implementation, sales commissions, and research-and-development costs remain unavailable. Consequently, no defensible gross-margin, contribution-margin, or cash-payback estimate can be built from the retained public evidence, and the unit-economics table treats those fields as diligence requests rather than zeroes.[CI002, CI014, CI015, CI016]

4.4 Capital adequacy, transaction change, and verdict

Third-party sources converge on $252 million raised historically, but funding raised is not cash on hand and says nothing about October 2026 burn, runway, debt, or contractual obligations. More importantly, an October 2026 aggregator reproduces an announcement that Harness acquired select Augment assets, including Cosmos, Auggie CLI, the Context Engine, and related technology, with the team joining Harness. The word “select” prevents treating this as proof that every asset or liability transferred, and no retained source provides consideration or balance-sheet treatment. The financial verdict is therefore not a stand-alone growth underwriting: pricing has a legible usage-based architecture, but revenue quality, margin, cash needs, and the post-transaction entity perimeter are unresolved. Diligence should prioritize the purchase agreement, closing statement, retained liabilities, customer-contract assignment, and a monthly cohort-level operating model.[CI005, CI009, CI010, CI011, CI015, CI016]

4.5 Exhibits

5.1 Product and user workflow

Augment Code is delivered inside an engineer's existing development workflow rather than as a standalone replacement for the IDE. The Visual Studio Marketplace listing describes a VS Code extension for professional engineers and large codebases, with Agent for completing tasks and solving production problems and Chat for questions, project planning, and defining changes. It also says Agent understands “your code, APIs, schemas, and dependencies,” while Smart Apply updates code in one click. The listed workflow is therefore broader than autocomplete: a developer can orient within an unfamiliar system, ask questions, plan changes, and apply multi-file edits. The same listing points users to Agent and Chat, native integrations, and “100+ tools with MCP.” This establishes the product surface and intended workflow, but not independent proof of the quality, latency, or safety of those operations.[CE001, CE002, CE003, CE004]

5.2 Technology architecture and maturity

Public descriptions consistently center the product on a Context Engine that maintains a representation of a large codebase and uses it to ground suggestions and agent actions. A Gartner Peer Insights page describes deep understanding of the large codebase, dependencies, and documentation, while third-party product summaries describe cross-file and cross-repository context and a Cosmos multi-agent layer. Those latter details should be treated as vendor-directory or market-description evidence rather than independently verified architecture. The available evidence does show a product that spans VS Code, agentic task execution, Chat, code review or PR workflows, and MCP-style integrations. A September 10, 2026 item in a news aggregator reports that Harness acquired select assets including Cosmos, Auggie CLI, and the Code Context Engine; without a primary announcement in the assigned pool, the transaction scope and resulting roadmap remain unverified. The evidence also does not establish the implementation of indexing, retrieval, orchestration, model routing, or multi-agent coordination, so the architecture map intentionally labels those as public descriptions and diligence items rather than verified internals.[CE005, CE006, CE007, CE008, CE013]

5.3 Trust, deployment, and implementation risks

The product's value depends on accurate context and safe application of changes across a customer's codebase, making integration quality, access boundaries, rollback, and reliability central diligence topics. Gartner user material reports strong contextual understanding and IDE integration, but also records occasional irrelevant suggestions, limited enterprise analytics, minimal offline support, and a need for manual intervention to avoid hallucination. A competitor-authored comparison additionally states that developers may encounter stale-file handling, reliability issues in migrations and integrations, and setup hurdles; because the issuer is a competitor, this is an adverse lead rather than a benchmark. The public pool contains no independent security audit, incident record, uptime history, or technical performance benchmark for Augment Code. Security and support controls, data retention, model-provider boundaries, deployment options, and production rollback procedures therefore remain material diligence gaps.[CE009, CE010, CE011, CE012]

5.4 Exhibits

6.1 Customer segments and buying motion

The clearest customer segment is the enterprise engineering organization rather than the individual developer seeking inexpensive autocomplete. WorkOS's April 2026 interview with Augment's CEO describes financial institutions, media companies, and a large furniture company building toward agent-heavy software engineering, while AIIDEList describes the product as best for enterprise engineering teams, large monorepos, production refactors, and repository-scale debugging. The buyer is therefore likely an engineering or technology organization with a complex repository and governance requirements; the end user is the engineer, technical lead, or reviewer who needs repository context. The public record does not disclose segment-level revenue, account counts, geography mix, or a verified payer profile, so those remain diligence questions rather than quantified conclusions.[CU001, CU002, CU003, CU004]

6.2 Adoption signals and durability

Public evidence shows several adoption surfaces but not durable cohort economics. The WorkOS interview provides a current qualitative signal that large, regulated and risk-aware enterprises are making infrastructure bets around agentic coding. Gartner Peer Insights presents a 4.8 rating from 42 ratings and includes reviews from software, telecom, and IT services employees; those reviews praise context-aware suggestions, codebase understanding, and code review while also noting occasional irrelevant or generic outputs and limited analytics. These signals are useful for product reception, but they do not prove active account retention, renewal, expansion, or production-wide deployment. Pricing and usage volatility are also relevant to durability: AIIDEList says usage-based billing can be harder to predict, and the WorkOS interview calls usage spiky, hard to predict, and growing fast.[CU005, CU006, CU007, CU008, CU009]

6.3 Named customer proof and concentration gaps

Named proof is uneven. The WorkOS interview identifies financial institutions, media companies, and a large furniture company only at category level, without names or deployment outcomes. The assigned directory sources describe enterprise use cases and integrations but do not independently establish production status, contract size, renewal, or expansion. Gartner's review page adds named reviewer contexts and dated user comments, which is stronger than a logo-only claim, but the record remains a small public sample and is not a customer census. Consequently, the report can support a thesis that Augment has credible enterprise adoption signals, not a claim that adoption is broad, retained, or concentrated in any particular account. Customer references, cohort retention, account-level usage, pilot-to- production conversion, and top-account exposure are the highest-value diligence asks.[CU010, CU011, CU012, CU013]

6.4 Exhibits

7.1 Severity-ranked risk posture

The highest-priority risk is execution through transition. A Harness-issued announcement reproduced by Parsers.vc says Harness acquired select Augment assets, including Cosmos, the Auggie CLI, and the Code Context Engine, and that the team behind those products would join Harness. That event makes product continuity, customer-contract assignment, roadmap ownership, and integration execution immediate diligence questions rather than hypothetical long-term risks. The next tier combines autonomous-action risk with commercial volatility: public user material describes terminal execution and database migrations, Gartner reviewers identify hallucination-related manual intervention, and independent reviews describe pricing changes and undisclosed credit-consumption rates. These signals do not establish measured incident probabilities, so likelihood labels below are qualitative and residual severity stays high where mitigation evidence is only reported rather than independently verified.[CR001, CR002, CR009, CR010, CR014, CR015]

7.2 Regulatory, legal, and control assurance

Secondary sources associate Augment with SOC 2 Type II, ISO/IEC 42001, customer-managed encryption keys, role-based access controls, audit logging, sandboxed execution, and no-training or zero-retention options. Those are relevant mitigations, but they do not establish the current audit period, scope, exceptions, shared-control obligations, subprocessor chain, or contractual remedies for a particular buyer. The assigned source pool also contains no primary litigation docket, regulator record, breach disclosure, or intellectual-property schedule. Silence in this pool must remain a gap, not evidence of a clean history. Legal approval should therefore require auditor-issued reports, the data-processing agreement, incident-notification terms, customer-data and model-training restrictions, indemnities, and confirmation of how the Harness transaction affects licenses and customer obligations.[CR001, CR007, CR008, CR010, CR011]

7.3 Operational, quality, and security exposure

Augment's product can act across code, terminals, databases, and connected development systems, so an inaccurate instruction or stale context can propagate beyond a wrong suggestion. A developer account describes the agent finding an execution error and attempting a fix, while Gartner's review summary says manual intervention can be needed to avoid hallucination. SiliconANGLE also reports Augment's own explanation that agents hallucinate and waste tokens on irrelevant context. Checkpoints, sandboxing, access controls, and audit logs are useful mitigations, but the pool supplies no independent uptime, rollback-time, incident-rate, penetration-test, or false-positive measurements. A production pilot should therefore use least privilege, mandatory review for consequential actions, representative failure injection, rollback tests, and monitored usage ceilings before autonomy expands.[CR002, CR003, CR005, CR008, CR009, CR010, CR011]

7.4 Partner, commercial, and financial dependencies

The asset acquisition concentrates near-term execution on Harness's integration decisions and creates uncertainty over product packaging, support ownership, and renewal terms. Technical dependencies are also broad: SiliconANGLE reports that the Context Engine can connect to external agents and model environments through MCP, while review sources describe dependence on IDEs and collaboration platforms. Commercial predictability is weaker. Vantaige reports two pricing-model changes in six months and cites one user's description of an effective increase approaching tenfold; AI Central Resources says credit-consumption rates and extra-credit pricing were undisclosed. CheckThat.ai reports no publicly disclosed revenue, growth, ARR/MRR, or acquisition metrics. Funding history is not cash availability, and no evidence here establishes burn, runway, margin, retention, customer concentration, or transition economics.[CR001, CR004, CR006, CR012, CR014, CR015, CR016]

7.5 Exhibits

8.1 Investment case and anti-thesis

Augment's positive case is built on an enterprise coding proposition, substantial institutional financing, favorable user-review evidence, and a reported strategic transaction. Tracxn describes a Palo Alto Series B provider of AI-powered coding-assistant and developer-productivity tools and reports $252M raised across three rounds. Gartner Peer Insights users describe contextual understanding and fast responses, while SiliconANGLE reported on October 8, 2026 that Harness acquired selected Augment assets, including Cosmos, Auggie CLI, the Code Context Engine, and related technology. That acquisition validates the strategic relevance of the product assets, but it does not disclose consideration, proceeds to shareholders, or the treatment of the remaining company. The anti-thesis is equally material: retained sources conflict on whether an ARR figure is publicly available, and none provides verified current growth, gross margin, retention, burn, runway, or acquisition economics. Parsers' archived April 2024 report of a $977M financing valuation is historical rather than a current mark. This combination makes the decision highly price-sensitive and disclosure-constrained.[CV001, CV002, CV003, CV004, CV005, CV007, CV014, CV016]

8.2 Entry discipline and scenario logic

A conventional revenue-multiple valuation cannot be responsibly calculated because no retained primary or high-confidence source verifies current ARR, growth, gross margin, net retention, or capitalization. CheckThat states that no public revenue, growth, ARR/MRR, or customer-acquisition metrics exist, while Vantaige reports approximately $20M ARR as of late 2025 without providing primary evidence in the fetched text. Tracxn reports $252M of cumulative funding and a latest Series B dated November 4, 2024, but masks that round's amount and valuation. Its competitor data provide funding comparisons - Sourcegraph at $223M, Tabnine at $102M, Codegen at $16.2M, and Magic AI at $466M - not revenue multiples or transaction valuations. The most defensible framework is therefore scenario-based: a bull case requires disclosed strategic consideration or verified recurring revenue and retention; a base case recognizes product value but no proof of shareholder proceeds; and a bear case allows limited common-equity recovery if the asset sale did not transfer the whole business. Entry discipline requires transaction documents, preference terms, financials, and customer concentration before assigning a price.[CV004, CV005, CV006, CV007, CV013, CV014, CV015]

8.3 Recommendation and thesis-break triggers

The current recommendation is research-more, with low confidence in any precise valuation stance and high risk because the central transaction and operating facts are missing or conflicted. The acquisition and favorable user reviews are positive validation signals, but neither establishes a purchase price, shareholder proceeds, or a verified recurring-revenue denominator. The 2024 financing history is useful historical context, not a current entry price. The thesis would strengthen if Harness or the transaction parties disclosed consideration, retained-business scope, customer and revenue transfer, or continuing capitalization; it would weaken if the deal was primarily an asset or talent acquisition with limited monetizable revenue, if customer contracts did not transfer, or if credit-based usage produced weak gross margins. Vantaige's report that pricing changed twice in six months adds monetization risk that should be tested rather than assumed away. Immediate diligence should obtain the definitive agreement, cap table and liquidation preferences, audited or management financials, cohort retention, customer concentration, and post-transaction product plans. Limited consideration, inability to verify recurring revenue, or absent durable customer contracts are thesis-break triggers.[CV002, CV003, CV008, CV009, CV011, CV012, CV013]

8.4 Exhibits