Webstack 2030
With the advent of CodeGen, the software stack is quickly becoming more declarative, semantic, and intent-driven. While the core primitives currently remain mostly consistent, our way of interfacing with them is evolving. We believe the biggest opportunity lies in reimagining how software is composed, orchastrated and delivered, with an emphasis on open systems and self-sovereign identity and data. Here are a few concepts we’re tracking:
1. Federated Runtimes & New Orchestration Protocols
A decentralized execution environment where code runs across federated nodes (devices, edge nodes, cloud), orchestrated by an open protocol. Think “Temporal meets IPFS.” Could support long-running workflows, LLM agents, and stateful coordination without centralized control.
Why now: As AI agents, edge computing, and real-time workflows become more common, traditional cloud runtimes are too centralized. This allows resilience, composability, and open participation.
⸻
2. Local-First Runtimes with Transparent SQLite Sync
A framework for building apps that run fully locally (mobile, desktop, edge) but sync seamlessly across devices and with servers when needed. Powered by SQLite + CRDTs or peer-based sync.
Why now: Privacy, offline support, and performance are rising priorities. Combine this with WASM or Bun/Deno for a compelling local runtime story. Perfect for productivity tools, field apps, and portable dev environments.
⸻
3. Universal Open-Source Auth with Self-Sovereign Identity
An Auth0/Clerk alternative built on decentralized identity (DID), passkeys, and user-owned data vaults. Developers get modern auth flows, but users own their identity and permissions across apps.
Why now: Identity is still fragmented, and centralized auth is increasingly a liability. The rise of wallets, passkeys, and zero-trust systems creates space for a new approach.
⸻
4. Composable Observability via LLM-Native Dashboards
An observability platform where devs describe what they want (“show me spikes in write latency by region”) and get generated dashboards, alerts, and remediation playbooks. Backed by OpenTelemetry, vector databases, and agentic interfaces.
Why now: Traditional observability is too noisy and reactive. This gives teams intuitive, proactive insight — and reduces dependency on manually configured dashboards.
⸻
5. Zero-Boilerplate Third-Party API Orchestration Layer
A tool that turns natural language into secure, version-controlled API integrations across SaaS tools. Think Retool + GPT + Zapier + Postman, but developer-first and CLI-compatible. With typed outputs, retry logic, and schema tracking.
Why now: Integrations are painful, repetitive, and fragile. LLMs and OpenAPI schemas make this ripe for abstraction. Could evolve into an AI-powered “integration runtime.”
⸻
6. LLM-First Workflow Engine for Human + Agent Collaboration
A new kind of orchestration engine built for mixed-initiative workflows (humans + agents). Think of it as “Airtable Automations meets LangChain meets Slack.” Declarative logic, memory, and feedback loops included.
Why now: LLMs need grounding, context, and human input. This creates a coordination layer for trust, iteration, and real-world execution — ideal for teams building agentic workflows.
⸻
7. Realtime Notification Fabric with Intent Understanding
A new notifications platform that’s intent-aware. Instead of “send email,” developers declare communication goals (“user must confirm their email within 1 day”), and the system handles channels, retries, timing, and copy using LLMs.
Why now: Notifications are a core UX primitive, but hardcoded and brittle. This abstracts delivery logic from intent — and can dynamically adapt based on user behavior, channel preference, or AI-tuned urgency.
⸻
8. Queryless Data Layer with Embedded Vector + Relational Stores
A new data abstraction where you don’t write queries—you express goals (“show me my most loyal customers”) and the system infers structure, joins, filters, and even embeddings. Think Prisma + SQLite + PGVector + GPT.
Why now: As LLMs increasingly understand data semantics, devs and analysts want intent-first access without writing SQL. Great for internal tools, dashboards, and agents.