Compound AI Systems and Orchestration

It’s Mid-Autumn Festival this week in China. Families gathering under the brightest moon, giving thanks for the harvest, praying for what’s to come. There’s something about that… the gathering, the coming together… that feels relevant to what’s happening in devtools right now (or am I just really trying hard to make a connection 😜). We’re seeing a shift. Not a loud one, but a clear one. The industry is moving away from the idea that one model solves everything. Instead, it’s about orchestration. Multiple specialized agents, models, tools… working together to solve complex problems.

The question being asked here isn’t “which model is best?” anymore. It’s “how do we make multiple models work together seamlessly?”

Just as the harvest moon symbolizes abundance through reunion, compound AI systems promise richer, more capable solutions through collaboration rather than competition. This week’s funding rounds, launches, and our DevTool of the Week all point to this… a fundamental shift in how we think about AI in production.

🏆 Funding Wins

Supernova Soars with $9.2M To Accelerate Product Workflows With AI

Czech startup Supernova has raised $9.2M Series A led by Taiwania Capital with J&T Ventures, Reflex Capital, Kaya, Credo Ventures, EQT Ventures, Wing VC and Y Combinator participating. Founded by Oskar Koristka and Jiri Trecak, Supernova also recently launched Portal, an AI-powered product development tool that spans ideation through to code. Built on Supernova’s unified platform for design systems and code, Portal introduces vibe-coding: using AI to manage routine and strategic tasks across product workflows. Unlike prototyping tools, Portal targets the entire product lifecycle and integrates with enterprise processes. Customers include Paramount, Mozilla and Skyscanner. The funds will grow the team, expand into Asian markets and scale product and marketing operations.

✨ Read the full announcement! Congrats Supernova!

Dash0 Secures $35M For AI-Native Observability

Dash0 has raised $35M Series A co-led by Accel and Cherry Ventures with DIG Ventures also participating. Founded by Mirko Novakovic and the team behind Instana, Dash0 is building the first AI-native observability platform with Agent0 at its core. Agent0 acts as an SRE copilot, helping developers fix application and infrastructure issues and create artifacts like alerts, dashboards and pipeline rules. Built on Dash0’s SIFT framework (Spam filter, Ingest, Filter, Triage) and OpenTelemetry standards, the platform has acquired over 270 customers in just nine months. Novakovic says:

“Observability today is broken: too noisy, too expensive, and too complex. … With SIFT, we’ve already built the foundation to remove noise, enrich signals, and surface root causes.”

The funds will accelerate US and European growth and double down on R&D.

🔍 Read the full story from the team.

Oneleet Raises $33M To End Compliance Theater

San Francisco startup Oneleet has raised $33M Series A led by Dawn Capital with Y Combinator, Dropbox co-founder Arash Ferdowsi and Frank Slootman also backing the round, bringing total raised to $35M. Founded by Bryan Onel, Ora Onel and Erik Vogelzang Oneleet consolidates penetration testing, code scanning, CSPM, attack surface monitoring, MDM and security training into one integrated platform. The company grew to eight-figure revenue profitably without touching its seed funding, all through organic growth. Onel and his team spent over a decade as penetration testers breaking into Fortune 500s and startups, repeatedly hearing: “We’re SOC 2 certified, ISO 27001 compliant, we spend millions on security… how did you get in so fast?” The answer: compliance became disconnected from security. Oneleet implements real security first, with compliance following automatically.

🔐 Brilliant news Oneleet! Learn more at Oneleet.

Modal Lands $87M To Pool The World’s Compute

Modal has raised $87M in Series B funding led by Lux Capital at a $1.1B valuation, bringing total raised to $111M. Founded by Erik Bernhardsson and Akshat Bubna, Modal built AI-native infrastructure that aggregates GPUs and CPUs globally in seconds, removing capacity headaches for developers. The platform runs on custom-built file systems, container runtimes and schedulers, delivering sub-second container starts and usage-based pricing. Customers like Perplexity, Lovable and OpenAI use Modal across inference, sandboxes, batch jobs, training clusters and notebooks. Meta recently used it to spin up thousands of concurrent environments for Code World Models reinforcement learning. The fresh funds will expand the product suite across the entire ML lifecycle.

🎯 Congratulations Modal! Read the founders’ announcement here.

Actual AI Announces $3.2M Seed Round To Automate Engineering Management

Seattle startup Actual AI has raised $3.2M seed funding led by AlleyCorp to build autonomous agents for engineering managers navigating the AI coding era. Founded by John Kennedy (CEO) and Ethan Byrd (CTO), the company targets teams of 50 to 500 developers where AI tools are creating new bottlenecks around review, coordination and maintenance. Actual AI’s agent automates recurring duties like triaging issues, generating sprint summaries, routing code reviews and enforcing architecture consistency. It also helps reduce reliance on senior engineers and upskill junior developers. Kennedy said:

“A lot of engineering managers want to be player-coaches… They want to be spending time with their engineers, up-leveling. They want to be helping with big, hard technical decisions. They want to be doing cool, fun stuff. They don’t want to be doing mechanical admin work.”

🤖 Amazing news for Engineering Managers! Explore Actual AI – they’re hiring too!

📈 Series E and Up

For the larger devtools players, there were some staggering raises this week:

• PostHog raised $75M Series E at a $1.4B valuation led by Peak XV with participation from Stripe, Y Combinator and GV, bringing total raised to $215M. 🦔 PostHog announcement

• Supabase raised $100M Series E at a $5B valuation led by Accel and Peak XV, 2.5x the valuation from its $200M Series D just four months ago. 🚀 Supabase announcement

• Vercel raised $300M Series F at a $9.3B valuation co-led by Accel and GIC. Vercel has doubled its user base over the past year with 82% top line growth. Its v0 agent has reached 3.5M users, with Teams and Enterprise now representing over 50% of v0 revenue.⚡ Vercel announcement

🚀 3, 2, 1… Launches

Verdent AI Launches Verdent Deck For Parallel AI Agent Coding

Founded by Zhijie Chen, Yuyu Zhang and Xiaochun Liu, Verdent AI launched Verdent Deck, a platform that coordinates multiple AI agents to tackle complex coding tasks in parallel with collision-free execution. The platform’s Plan Mode transforms rough ideas into structured plans with clarifying questions, then generates code reviewable via built-in Code Review and DiffLens tools. Verdent Deck is designed for professional engineers tackling large-scale tasks and delivers production-ready code. The company is also rolling out Verdent for VS Code with Verify and Research subagents for code checks and codebase understanding, plus autorun mode for automatic plan-to-execution. Verdent supports GPT-5 and Claude Sonnet 4.5, and hit the number 1 spot on Product Hunt last week.

🤖 Brilliant news Verdent! Get started at Verdent | Show support on Product Hunt

Benchify Launches Instant Self-Healing For Generated Code

Founded by Juan Castaño and Max von Hippel in 2024, Benchify (S24) tackles a core problem in vibe-coding: generated code breaks constantly with duplicate function calls, parse errors, stray tool calls and malformed diffs. Benchify sits between LLM clients and sandboxes, using static analysis and program synthesis to repair bugs before execution. It handles parsing, dependency, CSS, Tailwind, type and interaction errors, then bundles code in 1-3 seconds. For front-end work, developers can skip sandboxes entirely and render directly from Benchify’s output. The company is hiring two founding engineers, one in SF and one in NY.

🔧 Amazing news for app builders! Explore Benchify | Product Demo

Continue Launch Week With Notch Redesign And Custom Agents

Continue launched a week of updates bringing continuous AI to every developer and team.

• Day 1 introduced Notch, a redesigned VS Code interface with intuitive UI and intelligent context management.

• Day 2 launched custom AI agents and community cookbooks for automating code reviews, test generation and documentation.

• Day 3 released a PostHog Analytics Agent cookbook for querying analytics in natural language.

• Day 4 introduced Workflows, background agents that run continuously to address bugs and build boilerplate features.

• Day 5 opened documentation for community contributions via Hacktoberfest.

🎉 Love it Continue! Find out more and keep up-to-date with this releases here – they’re hiring too!

Merge Launches Agent Handler For Secure AI Agent Integrations

Merge launched Agent Handler, a platform that securely connects AI agents to enterprise-ready tools via MCP with built-in auth, credential management and security. Merge already powers customer-facing integrations for frontier LLMs, Fortune 500s and B2B SaaS with two products: Merge Unified (hundreds of integrations with one API) and Merge Agent Handler (secure access to thousands of third-party tools). With enterprise-grade security and full integration lifecycle management, thousands of companies trust Merge to speed development, unblock sales and cut churn.

🔗 Congrats Merge! Try Agent Handler free | Find out on Product Hunt what Product Marketing Manager, Irene Hu has to say.

Tiptap Pages Brings Document Structure To Open-Source Editors

Founded by Philip Isik, Tiptap launched Tiptap Pages into live beta last week, adding pagination, page layouts, headers, footers and structure to the open-source text editor framework used by thousands of dev teams. Pages solves a core problem: most rich text editors treat documents as one long scroll, which doesn’t work for contracts, RFPs, government proposals or legal docs where structure matters. Pages brings visual page layouts (A4, Letter, Legal or custom), smart pagination, dynamic headers and footers, and custom node support.

📄 Well done Tiptap! Try the live demo and find out more on their website.

Velt Launches Multiplayer Editing Toolkit For React Flow

Founded by Rakesh Goyal, Velt launched a purpose-built plugin for React Flow that makes AI agent workflows and workflow builders multiplayer in minutes. The toolkit handles fully managed ultra low latency backend, real-time merges and conflict resolution via CRDT (Yjs), offline support, presence indicators, Slack-style huddle, Figma-style comments and notifications, and custom encryption. Developers can focus on delivering value without building complex infrastructure from scratch. Velt powers multiplayer experiences for companies like Heygen, Pendo, Runway and Vellum.

🕸️ Brilliant news for AI workflow builders! Try Velt today.

Fern Launches Swift SDK Generator For iOS Developers

Fern launched a Swift SDK generator that instantly provides iOS developers with an idiomatic Swift client for any API. The generator supports REST (OpenAPI), WebSockets (AsyncAPI) and gRPC (Protobuf) APIs. Fern’s Swift SDKs feature zero external dependencies, modern async/await support, strong typing, forward-compatible structs, and automatic retries with exponential backoff. Unlike Apple’s Swift SDK generator which outputs files with over 40,000 lines of code, Fern breaks SDKs into multiple files with minimal namespacing. Early design partner Aaron Carver from Chrt said:

“Using our API from Swift is basically frictionless with Fern’s SDK generator.”

🌿 Get started with the Swift Quickstart | Read reviews on Product Hunt – great work Fern!

 

🔦 DevTool of the Week: Fireworks AI

Founded in 2022 by a team of Meta AI veterans, Fireworks AI has rapidly become one of the most compelling platforms in the generative AI space. CEO Lin Qiao , who previously led the PyTorch team at Meta, co-founded the company alongside Dmytro Ivchenko, Benny Yufei Chen, Chenyu Zhao, Dmytro Dzhulgakov (CTO), James Reed and Pawel Garbacki.

Together, they’ve built a generative AI platform that processes over 10 trillion tokens daily with 99.99% API uptime, serving developers at Cursor, DoorDash, Quora, Upwork and beyond.

What caught our attention this week is the sheer pace of progress. In July 2024, Fireworks raised $52M Series B led by Sequoia Capital, boosting its valuation to $552M and bringing total raised to $77M. Investors include NVIDIA, AMD, MongoDB Ventures, Benchmark, Databricks Ventures, and former Snowflake CEO Frank Slootman. But it’s what they’ve shipped since that really stands out.

Last month, Fireworks launched Streaming Transcription V2 and Streaming Speaker Diarization for real-time speech-to-text and live speaker tracking. Just last week, they announced Fireworks for Startups, a program providing AI-native startups with platform access, tools and expertise to accelerate time to market. They’ve also released Document Inlining for reasoning over text and images, and continue to push the boundaries of compound AI systems, an approach that disaggregates tasks into sub-tasks allocated to the best model for the job.

In a recent interview on The Stack, Lin describes their philosophy simply:

“The next wave of quality is not going to be one of ‘single model solves all problems.’ There’ll be hundreds of small expert models solving narrower sets of problems.”

Fireworks abstracts the complexity away, delivering fast, affordable, highly performant generative AI capabilities across multiple modalities with an eye on solving the industry’s three main problems: cost, quality and latency. Their infrastructure offers three deployment options (serverless, on-demand, and enterprise-ready BYOC) and runs on NVIDIA H100 and A100 GPUs through AWS.

Deep partnerships with MongoDB, Meta, NVIDIA, AMD and AWS provide developers with fundamental building blocks from state-of-the-art models to vector databases on a production-ready platform.

In a world where keeping up with generative AI innovation feels impossible, Fireworks AI is doing the heavy lifting so developers can focus on what matters: building groundbreaking products.

🚀 Explore Fireworks AI | Join Fireworks for Startups

Got a launch, raise, or tool we should feature next week?

Reach out to one of our team.

Louise Ogilvy (Recruitment Director) – UK/USA

Becca Combe MIRP CertRP (Lead Recruiter) – USA

Ben Chipchase (Lead Recruiter) – UK

Natalie Harper (Marketing Administrator)

Looking for your next DevTools role?

Visit devtoolsjobs.com