Mango Capital

Deployment has been the last bastion of human judgment. That's about to change. At thousands of changes a day, a human clicking "approve" becomes the throughput ceiling of the whole factory. So the human role moves up: engineering leadership stops looking like standups and tickets, and starts looking like running a trading desk — setting policy, allocating budget, tuning the controls. Part 4 is the final layer: autonomous deployment, the control plane that governs it, and the agents — release, SRE, cost, compliance, migration — that run the factory. We backed env zero and Serro at inception. And it closes on a moonshot — a Code Genome Project: mapping the world's code the way biology mapped the genome, built on the lossless code representations Moderne has pioneered. #SoftwareFactory

You can't inspect quality in at the end. Toyota proved that in the 1980s. Deming said it for thirty years. Software never learned it — and agents writing code at 100x velocity are about to force the bill. You can't review your way to trust. Quality has to be engineered into the process, not caught at the end of it. Part 3 is the Six Sigma software factory: testing as a co-producer of code, continuous threat modeling, SLAs as engineered tolerances, cost as a first-class signal. We backed Checkly, Heal.dev, ProjectDiscovery, Perfloop, and Infracost at inception because the validation layer can't be an afterthought. https://lnkd.in/e75zHYw7 #SoftwareFactory

Gartner just created the category. Moderne is a Leader in it. We seeded Jonathan Schneider and Olga Kundzich in 2021, back when this wasn't a market — it was an open source project called OpenRewrite. Congratulations to the whole Moderne team.

Turso is building the LLVM of databases — one engine, in Rust, that covers many database services including: SQLite, Postgres, document, vector, views, mvcc, browser and more. Postgres brings the ecosystem. Turso brings the density and efficiency agents require: 1M+ databases created per week for a single customer! Congratulations Glauber Costa and Pekka Enberg. Mango Capital has been in since inception. #ConvictionBeforeConsensus #AIInfrastructure

AI agents are only as useful as the web they can act on. This month TinyFish landed three partnerships — Glean, Alibaba Cloud, and Amazon Web Services (AWS) Marketplace — putting live-web infrastructure everywhere agents already work. Congrats to Sudheesh Nair,  Shuhao Zhang, and the team — details in the comments. #ConvictionBeforeConsensus #AIInfrastructure

10 years of NetBox. 20k GitHub stars. 4x YoY growth. Today NetBox Labs launches its biggest update yet — an Infrastructure Intelligence Platform for humans and agents. Congrats to Kris Beevers and the team — details in the comments. #ConvictionBeforeConsensus #AIInfrastructure

Generating code is only half the battle. A factory needs a warehouse, inventory, and logistics. Software is about to discover it has none of these. Agents committing at machine velocity breaks everything underneath — Git, artifact repos, Kubernetes. Built for humans. Not for this. Part 2 is the generational rebuild: source control for agent-native access, artifact repos as provenance graphs, runtimes that collapse build loops from minutes to milliseconds. We backed Unikraft at inception because we saw the need for more granular compute. https://lnkd.in/exp8a4eV #SoftwareFactory

Agents don't just move data — they sense, decide, coordinate, and act. That demands streaming infrastructure built to perform under pressure. Mango Capital invests with conviction before consensus, and LaserData is exactly that bet: sub-millisecond, Rust-native streaming, native agent-protocol support, enhanced data services, and a free tier live now on AWS and GCP. Congrats to the LaserData team — full details in the comments. #ConvictionBeforeConsensus #AgentInfrastructure

Software development is about to hit an industrial revolution. As coding agents make developers exponentially more productive, the constraint shifts from writing code to absorbing it. At exabyte-scale volume, the artisanal model of software engineering — and every system built around it — breaks. Technical debt, security risk, and maintenance overhead don't scale linearly; they buckle. The answer isn't better tools for individual craftspeople. It's the software factory: governed, repeatable, self-learning processes running at machine scale. Part 1 starts at the coding layer — and the context problem that makes or breaks agents in real enterprise environments. Mango Capital backed both Moderne and Upstash / Context7 at inception, and runs Bauhaus, a dedicated seed program with Moderne, to build toward it. Full piece in the comments. #ConvictionBeforeConsensus #SoftwareFactory