HDT
Start a build

Software is the raw material.

We build software across whatever domain the problem lives in: agents, real-time 3D, data systems, products, automation. Everything you see behind this text is running live on your GPU, made of the same thing we ship.

Scroll to compile

Every artifact begins as text.

The lines of code you just scrolled past have been compiled into the object on the right. Its shape is set by the parameters in this editor. Drag any orange number (or focus it and use the arrow keys) and roughly fifty thousand vertices re-form on the GPU in the same frame.

forge.ts — illustrative
import { compile } from "@hdt/forge";
 
const artifact = compile(source, {
  frequency:  1.60,
  amplitude:  0.34,
  twist:      0.80,
  turbulence: 0.22,
  heat:       0.65,
});
 
artifact.render(world); // you are looking at it
-- vertices · frame -- ms

Then it has to survive the real world.

Beautiful is easy for a week. We engineer for what happens next: traffic spikes, dead nodes, slow dependencies. Below is a small distributed system simulated in your browser. Surge the traffic, click any node to kill it, and watch it reroute, queue and scale.

Throughput / s
0
p95 transit
0.00s
Dropped
0
Workers
3

    Simulation. Time is compressed; numbers describe this model, not a client system.

    One company. No fixed terrain.

    We don't pick a niche and wait for it to fit the problem. We go where the problem is. Recently, that has been here:

    1. /01

      Autonomous agents

      Multi-agent systems with memory, tools and guardrails that do real work, not demos. Deployed orchestrations managing automated research, triage, and multi-model synthesis pipelines.

      orchestration · retrieval · evals
    2. /02

      Real-time graphics

      WebGL, shaders and 3D experiences, like the one under this page. High-density GPU particle systems, custom post-processing pipelines, and 60 FPS interactive visualizers.

      three.js · glsl · webgpu
    3. /03

      Data & infrastructure

      Pipelines, APIs and back-ends sized for the load they will actually see. Sub-10ms transactional endpoints, resilient queue fabrics, and zero-loss streaming architectures.

      postgres · queues · observability
    4. /04

      Product software

      Web and mobile applications with a studio's craft and a platform team's discipline. Full-lifecycle systems with automated regression test gates and sub-second cold starts.

      typescript · react · native
    5. /05

      Automation

      Internal tools and workflows that turn hours of manual operations into a scheduled job. Resilient headless scrapers, sync jobs, and webhook-driven event engines.

      integrations · scraping · ops
    6. /06

      Developer tooling

      Code indexes, CLIs and registries that make other engineers faster. AST symbol graphs, capability catalogs, and CI linting harnesses deployed across client codebases.

      dx · cli · static analysis

    Five passes from vague to running.

    PASS 01

    Probe

    We read the problem until it stops being vague: constraints, data, users, and the failure modes nobody mentioned.

    Spec: constraint matrix · failure modes · schema boundary
    PASS 02

    Prototype

    Something real, fast. A working spike settles arguments a slide deck can only start.

    Artifact: vertical spike in production runtime
    PASS 03

    Harden

    Types, tests, load, observability. The unglamorous part, done on purpose and early.

    Verification: invariant fuzzing · chaos checks · p99 SLAs
    PASS 04

    Ship

    Automated pipelines, staged rollouts, deploys you can reverse in one command.

    Delivery: progressive canary · single-command rollback
    PASS 05

    Operate

    Software is alive after launch. We instrument it, watch it, and keep it evolving.

    Telemetry: distributed tracing · SLO alarms · auto-healing

    Bring us the one that sounds impossible.

    Tell us what you want to exist. We'll reply with how we would build it.

    Start a build