AI application-layer startup radar: August 5–12, 2026

Executive read

The August 5–12 window produced 14 application-layer financing disclosures with stated amounts: nine dollar rounds totaling $570.1 million, plus five euro rounds totaling €22.71 million. One additional follow-on for Platter had no amount disclosed. These are current-round figures only; USD and EUR remain separate rather than being converted at an arbitrary exchange rate.

The biggest number is Lovable's $400 million Series C at a $13.3 billion valuation. It is a capital-market signal from a mature company, not evidence that Lovable first appeared this week. The more useful early-stage map is broader: Malachyte is moving recommendation systems from historical profiles to live shopping intent; Relu is turning dental scans into treatment designs; Zerolook is predicting flight itineraries before an agent generates expensive searches; and STRGY is attaching AI to the daily execution of corporate strategy.

A second cluster is forming around control after deployment. FriskAI records what agents do in production, Palette keeps company context durable while models change, Cytix manages change risk in AI-accelerated software development, and Blacksmith validates the code those systems generate. The common diligence question is no longer just whether a model can complete a task. It is whether the startup owns the data, permissions, workflow state, and audit trail that make the task safe to run repeatedly.

Quick scan

Company	Sector / status	Window funding	Product positioning	Founding-team signal
Naïve	AI agents / boundary case	$28.5M Series A; Nexus Venture Partners; Aug. 6	API for payments, email, phone, cloud, storage, incorporation, and agent operations	Sean Dorje is named CEO and co-founder; no prior affiliation disclosed
Malachyte	Vertical SaaS / e-commerce	$10M seed; Bessemer Venture Partners and Gradient; Aug. 6	Real-time, intent-aware product recommendations for online stores	Sidd Motwani, Ian Anderson, and Shivaditya Sinha built Spotify recommendation infrastructure
Omilia	AI agents / mature company	$67M Series B; Expedition Growth Capital; Aug. 6	Self-learning agents for voice and customer-support workflows	Dimitris Vassos is named CEO; fuller founding history not disclosed
Relu	Vertical SaaS / dental	€1.51M strategic investment; Dental Innovation Alliance; Aug. 6	Automated design and planning from 3D dental scans	Holger Willems, Adriaan Van Gerven, Antoine Coppens, and Thomas Beznik; prior affiliations not disclosed
PineGap.ai	Vertical SaaS / finance	$8M Series A; Stellaris Venture Partners; Aug. 7	Custom AI agents for institutional equity research	Ankit Varmani is a former JPMorgan analyst; Deepak Sharma is an IIT engineer
Platter	Vertical SaaS / food supply	Follow-on; amount undisclosed; Verb Ventures; Aug. 7	Food-native operating system for orders, production, inventory, and finance	Founders not disclosed on the opened source
Echovane	AI agents / market research	$1M pre-seed; Titan Capital and Neon Fund; Aug. 10	Agents handle research design, recruitment, fieldwork, and synthesis end to end	Smriti Gupta, Vipul Nair, and Himadri Roy previously worked at Amazon, Stripe, and Razorpay
FriskAI	AI agents / observability	$3.6M pre-seed; MaC Venture Capital; Aug. 11	Runtime intelligence that records agent tools, arguments, responses, and timing	Neel Palrecha is named founder and CEO; prior affiliation not disclosed
Palette	AI agents / team operating system	€3M pre-seed; Ugly Duckling Ventures; Aug. 11	Durable company context and swappable models for AI-native teams	Brian Kyed, Lars Ettrup, Steffen D. Sommer, and Christian Lomholt; prior affiliations not disclosed
Quartr	Vertical SaaS / finance data	€15.6M; Altos Ventures; Aug. 11	First-party investor-relations data and AI research tools	Oscar Küntzel, David Dag, Oliver Hamrin, and Sami Osman; prior affiliations not disclosed
Zerolook	Vertical SaaS / travel	€1.6M pre-seed; Playfair; Aug. 11	Predictive flight-shopping API for airlines, agencies, and AI-generated search demand	Simone Lini previously led Google Travel partnerships and founded Waynaut; George Hadjiyiannis was VP of product at Kayak
STRGY AI	Vertical SaaS / strategy	€1M angel round; private investors and Innovestor Angel CoFund; Aug. 12	StrategyOS connects daily work to strategic priorities and flags drift	Samuli Bäck and Anton Skarp are co-founders; Oskari Listomaa and Niko Savander lead agentic systems and design
Blacksmith	AI agents / developer workflow; mature company	$45M Series B; Peak XV Partners; Aug. 12	Continuous integration, testing, and an agent that fixes failed code checks	Aditya Jayaprakash is named co-founder and CEO; no fuller background disclosed
Lovable	Consumer AI / mature company	$400M Series C at $13.3B; Menlo Ventures and Scaleup Europe Fund; Aug. 12	Software-creation platform for building and running businesses	The official announcement names founders Anton and Fabian but gives no biographies
Cytix	Vertical SaaS / security	$7M Series A; Northern Gritstone; Aug. 12	Change-risk management between the software lifecycle and security/compliance teams	Founder background not disclosed on the opened source

A blank biography is a source boundary, not an invitation to fill the field from an unverified database. Traction figures below are company-reported unless the source says otherwise.

The early-stage signals

Malachyte: recommendation becomes a live intent system

Malachyte raised a $10 million seed round co-led by Bessemer Venture Partners and Gradient, with Harpoon Ventures participating. The company was founded by Sidd Motwani, Ian Anderson, and Shivaditya Sinha, who spent years building the behavioral-intelligence infrastructure behind Spotify's recommendation engine. TechCrunch says that system powers about 90% of Spotify's recommendations to its 800 million users. 1

The e-commerce product is not just a better catalog recommender. It predicts both a shopper's general taste and what that person is trying to accomplish in the current session, using signals such as searches, clicks, scrolls, and add-to-cart actions. The platform went live with Fun.com in fall 2025 and became generally available to Shopify merchants through a native integration in June 2026; larger retailers can use its API. The company says it has worked with more than 20 enterprise customers across travel, grocery, and retail. 1

The diligence hinge is data density inside a session. If the system improves as a shopper acts, Malachyte may own a useful intent layer that first-party stores have not operationalized. If the signal is not strong enough to change conversion or basket size, it is another personalization feature competing with platform-native tooling.

Relu: regulated workflow depth, not a dental chatbot

Leuven-based Relu announced a €1.51 million strategic investment from Dental Innovation Alliance. Its Relu Automate platform analyzes 3D scans and can generate designs such as night guards, surgical guides, and crowns. The broader treatment workflow combines CBCT segmentation, intraoral-scan alignment, implant positioning, abutment positioning, provisional crown design, and guide design. 2

Relu says its software handles more than one million cases a year across dental laboratories in 82 countries and has 33 peer-reviewed publications. It also reports ISO 13485 certification, FDA 510(k) clearance, and CE marking. Holger Willems, Adriaan Van Gerven, Antoine Coppens, and Thomas Beznik are named as co-founders, but the opened source does not give their prior affiliations. 2

For diligence, the important asset is the case workflow and its regulatory evidence. A model that produces a plausible design is easy to demo. A system that fits a certified, repeatable, pay-per-case production process is harder to displace. The open question is how much of the reported case volume flows through Relu's paid product rather than through broader platform exposure.

PineGap.ai: research agents with a buyer and a format

PineGap.ai raised an $8 million Series A led by Stellaris Venture Partners, with Inventus Capital Partners, Silicon Valley Quad, and DeVC participating. The New York- and Bangalore-based company lets buy-side firms deploy AI agents configured around each institution's research methodology, proprietary datasets, and reporting formats. 3

The company says it has worked with more than 100 institutional buy-side funds, deployed over 1,000 agents, and produces more than 50,000 research reports per month. Those are company-reported figures. Founder Ankit Varmani previously worked as a JPMorgan analyst and spent eight years at a $2 billion long-short equity hedge fund; co-founder Deepak Sharma is described as an engineer from the Indian Institute of Technology. The technology organization is led by founding research engineer and head of AI Avijit Thawani, who holds a PhD from the University of Southern California. 3

PineGap's defensibility claim is more specific than generic "AI analyst" positioning: the agent is trained around the customer's own method and output constraints. The buyer-side test is whether those configurations improve coverage and decision speed without creating unacceptable hallucination or compliance risk. Quartr, covered below, is pursuing a related finance workflow from the first-party data side.

Echovane: a research service packaged as an agent system

Echovane raised a $1 million pre-seed co-led by Titan Capital and Neon Fund. Its agents cover market research from study design and participant recruitment through multimodal fieldwork and synthesis of qualitative and quantitative data. The company says its customers include Coca-Cola, Procter & Gamble, Haleon, Kantar, and Trustly; a Trustly executive said Echovane compressed a month of product research into two days. 4

Smriti Gupta, Vipul Nair, and Himadri Roy previously worked on product development at companies including Amazon, Stripe, and Razorpay. That background fits the company's attempt to own the research outcome rather than sell a single interview or survey tool. 4

The key diligence question is verification. Recruitment quality, sample bias, and the difference between a fast answer and a reliable answer determine whether the product can move from pilot budget to recurring research spend. A faster synthesis layer is useful; a trusted research process is the moat.

FriskAI and Palette: two answers to the same deployment problem

FriskAI launched with a $3.6 million pre-seed led by MaC Venture Capital, with Wischoff Ventures, NEA partner Rick Yang, and angels participating. Its runtime-intelligence platform logs agent tool usage, arguments, responses, and timing, then builds behavioral profiles and flags deviations across versions and deployments. Health benefits provider Sana Benefits is described as an early user. Founder and CEO Neel Palrecha is the only named team member, with no prior affiliation disclosed. 5

Palette raised a €3 million pre-seed led by Ugly Duckling Ventures, with Emblem, Acadian Ventures, and angels participating. Its thesis is different: keep a company's context, workflows, and knowledge durable while allowing teams to swap models and providers. Palette Desktop is available to paying customers, while the broader Palette OS is being used by design partners. Brian Kyed, Lars Ettrup, Steffen D. Sommer, and Christian Lomholt are named as founders; the source gives no previous affiliations. 6

Together they define two adjacent control points. FriskAI asks, "What did the agent actually do?" Palette asks, "What context and process should survive when the model changes?" Both need to prove that control becomes part of the customer's operating system rather than a temporary layer added during an AI pilot.

Zerolook: predict the search before the agent creates the cost

Zerolook raised a €1.6 million pre-seed led by Playfair, with Vento Ventures, TrueSight Ventures, Alpha Venture, and angels participating. Its B2B flight-shopping API predicts itineraries and prices instead of computing every possible search, targeting airlines and travel agencies that may face a new wave of demand from AI agents. The company says it has signed letters of intent with online travel partners ahead of first pilots. 7

The founding team supplies unusually direct distribution and domain evidence. CEO and co-founder Simone Lini previously led Google Travel's OTA and metasearch partnerships across EMEA, founded Waynaut before its acquisition by Lastminute.com, and later became Lastminute.com's chief commercial and product officer. Co-founder and CTO George Hadjiyiannis was VP of product at Kayak and has worked across Google, NVIDIA, and eBay. 7

The product's proof point is not a travel chatbot. It is lower marginal search cost and sufficiently accurate price prediction for a partner to trust the API. Letters of intent are an early distribution signal, not evidence of production volume.

STRGY AI: an operating system for strategy drift

Helsinki-based STRGY AI announced €1 million in angel funding from private investors across the UK, Norway, Switzerland, and Finland, with additional equity participation from Innovestor's Angel CoFund and non-dilutive support from Business Finland. Its StrategyOS gives Chiefs of Staff, Heads of Strategy, and COOs an always-on view of how daily work connects to strategic priorities, flags drift, and automates board-ready reporting. The company says the platform is already used by early customers in consumer brands and private-equity-backed companies, with more enterprise deployments in progress. 8

Samuli Bäck is co-founder and CEO, Anton Skarp is a co-founder, and Oskari Listomaa and Niko Savander are identified as the heads of agentic systems and agentic design. The source does not disclose previous affiliations. 8

The wedge is a familiar management problem with a measurable failure mode: priorities drift between planning meetings. The diligence test is whether StrategyOS sees enough underlying work to detect drift earlier than existing project-management and business-intelligence systems, and whether leaders will pay for the alert rather than merely read it.

Mature-company financing signals

Lovable: software creation becomes a consumer-scale application layer

Lovable announced a $400 million Series C at a $13.3 billion valuation, led by Menlo Ventures and the Scaleup Europe Fund, managed by EQT. The official announcement lists Balderton, Carmignac, Kaszek, LTS Growth, Tencent, World Innovation Lab, Regent, Accel, Antler, CapitalG, DST Global, Evantic Capital, HubSpot Ventures, and Salesforce Ventures among the participating investors. 9

Lovable says users have created more than 60 million projects since its November 2024 launch and that those apps receive more than 900 million visits per month. TechCrunch separately reports a $500 million annualized run-rate revenue figure disclosed by the startup in June. These are company-reported metrics. 10

This round belongs in the radar because it shows where application-layer capital is concentrating: a tool that lets non-traditional software builders create and operate businesses. It should not be read as a newly emerged company in this week's map. The diligence question has shifted from whether vibe coding works to whether Lovable can retain users after the first prototype, when security, maintenance, distribution, and operating cost become the real work.

Omilia: the ROI case for older customer-support infrastructure

Omilia raised a $67 million Series B led by Expedition Growth Capital. The Athens-based company has worked on voice automation and customer support since 2002, and it says its annual recurring revenue has grown tenfold to $60 million since its previous financing in 2020. Its customers include Capital One, Discover, RBC, DWP, PSEG, and Taco Bell; the company says its technology is deployed across more than 1,000 Taco Bell outlets. 11

CEO Dimitris Vassos says the company uses multiple tools rather than forcing every support task through a large language model. That is the useful signal: application-layer economics may favor a mixed system when the workflow contains many routine, structured requests. The counter-signal is deployment reliability. A customer-support agent that handles a narrow request correctly has a different risk profile from one that improvises across a whole contact center.

Blacksmith: code generation creates a validation budget

Blacksmith raised a $45 million Series B led by Peak XV Partners, with GV and Y Combinator participating, at a $550 million valuation. The company says it now serves more than 5,000 customers, including Mercury, Supabase, Clerk, Ashby, and Expensify. 12

Blacksmith began as cloud infrastructure for continuous-integration workloads and expanded into Codesmith, an AI coding agent that can fix failed checks. Co-founder and CEO Aditya Jayaprakash says the company reached a $10 million annualized run rate with 10 employees and now reports revenue in the tens of millions, without giving a current ARR number. These figures are company-reported. 12

The competitive set includes GitHub Actions, Cursor Automations, model-native validation features, and cloud providers. Blacksmith's application-layer opportunity is the budget created when AI makes code cheap to generate but expensive to verify. Speed, price, and a trusted history of validation will matter more than another code-generation demo.

Cluster notes for diligence

1. The application layer is moving toward system-of-record ownership

Malachyte reads live shopping intent. PineGap stores institution-specific research methods. Quartr is building a structured first-party data layer for public-company research. Zerolook sits between AI-generated travel demand and airline inventory. Relu sits inside a regulated dental production workflow. Platter describes itself as a system of record for food trade, while STRGY wants to become the record of strategy execution.

These companies are not interchangeable. Their shared signal is that the model is only one component. The defensibility question is who owns the durable state: the customer profile, the research method, the flight-pricing relationship, the clinical case, the inventory ledger, or the work graph. A generic model vendor can add a feature; it cannot automatically acquire the customer's permissions, historical decisions, and operating relationships.

2. Agent adoption creates a control category

FriskAI, Palette, Cytix, Blacksmith, Naïve, and Omilia attack different failure points around deployed automation. FriskAI provides runtime evidence. Palette preserves context across model changes. Cytix assesses change risk. Blacksmith validates generated software. Naïve packages the business systems an agent needs to operate. Omilia demonstrates the older version of the same thesis in a customer-support environment.

For a VC associate or strategy team, the comparable fields are deployment time, integration depth, auditability, and gross-margin exposure. A product that needs bespoke data cleanup for every customer may be selling a services project with a software wrapper. A product that converts each implementation into reusable schemas, evaluation data, and permissions can build a compounding asset.

3. Capital is rewarding both new wedges and proven distribution

The dollar total is dominated by Lovable's $400 million round, followed by Omilia's $67 million and Blacksmith's $45 million. The euro total is led by Quartr's €15.6 million. That mix makes the week's headline financing look larger than the early-stage map actually is.

The early-stage entries are more useful for finding new company formation: Malachyte, Relu, PineGap.ai, Echovane, FriskAI, Palette, Zerolook, and STRGY AI all attach AI to a specific buyer, dataset, or workflow. Their next diligence step is not valuation comparison. It is a reference call or product trial that tests whether the claimed proprietary surface changes an operating metric.

What to carry into deeper diligence

Workflow ownership: Does the product replace a manual handoff, or add another dashboard beside the system that still owns the decision?
Proprietary data: What gets better with every customer action, case, search, or deployment, and can the customer export or reproduce it?
Distribution: Is there a built-in channel, such as Shopify, Google Travel, a regulated partner, or an existing system of record?
Deployment effort: How long does a customer take to reach reliable production, and how much work remains bespoke?
Auditability: Can a buyer reconstruct what the agent saw, decided, changed, and failed to do?
Defensibility: Which part survives a model vendor adding the same interface next quarter?

Sourcing and scope notes

Window: public disclosures dated August 5–12, 2026, in the channel's Pacific-time schedule. The opened sources generally provide a calendar date or a publication timestamp; where only a date is available, this issue makes no hour-level cutoff claim.
Counting: the headline total includes nine dollar rounds with stated amounts and five euro rounds with stated amounts. Platter's undisclosed follow-on is described but excluded from both subtotals. Prior rounds and cumulative funding are context, not additions to the current-round total.
Classification: Naïve is included as a boundary case because its product automates business operations for agents, while its API and infrastructure language sits close to the infra boundary. Model-only companies and general compute providers are excluded. Mature-company rounds are labeled as capital-market signals rather than new entrants.
Evidence: each funding, product, traction, and founder claim is attached to the opened original article or company announcement. Company-reported customer, ARR, usage, and case-volume figures are labeled as such; missing founder backgrounds remain unavailable.

The next issue will continue separating newly surfaced application companies from follow-on financings, with particular attention to whether this week's control and workflow products show repeatable deployments rather than strong launch narratives.