What's been shipped.

Full xterm.js terminal embedded in the dashboard. Select text, Ctrl+C / Ctrl+V, right-click paste — all working.

GPT-4o and GPT-4o mini now route to the live OpenAI API. Streaming responses, proper token counting, accurate quota errors surfaced to the UI.

AI responses now render headings, bold, code blocks, lists, and inline code correctly instead of raw markdown strings.

Shipped the core dashboard with Research, Biology, Flywheel, and Write modes — each with a tailored system prompt and UI.

Model switcher in the dashboard header. Each model streams responses; DeepSeek R1 shows chain-of-thought reasoning traces.

Connect GitHub, HuggingFace, W&B, Modal, Pinecone, and OpenAlex directly from the dashboard. Credentials stay in-browser.

CPU free tier plus GPU profiles at ₹12/hr (T4), ₹45/hr (A10), and ₹150/hr (A100). Select before any training run.

Run automated benchmarks against custom metrics. Results logged per run, comparable across model versions.

All new accounts get one free message per model — no credit card, no friction. Enough to see the difference between modes.

Full authentication flow: create account, sign in with password, or get a one-click magic link via email. JWT sessions, Bearer token to API routes.

Launched the pricing page with three tiers. Early Access locked at ₹199/month — permanently. Waitlist modal collects emails.

Nav collapses on small screens, grid sections reflow to single column, touch targets sized correctly throughout.

First public version. Landing page with the Veil concept, neural canvas background animation, and an email waitlist.

Research, Biology, Flywheel, and Write modes defined. UI mockups and mode descriptions published on the landing page.