Loading…
Long-form posts from our engineers, designers, and strategists. No thought leadership — just the things we wish we'd known on day one.
Anthropic's Opus 4.7 just shipped with a 1M token context window and 12-hour autonomous task budgets. Here's what changes for product teams in 2026.
AIDeepSeek-R2, Qwen3-Think, and the 8B reasoning class are doing 70% of the work for 5% of the cost. The architecture change you need to make.
Read
EngineeringApple's second-gen headset shipped lighter, cheaper, and with proper enterprise MDM. After six pilots, here's where it actually earns its keep — and where it doesn't.
Read
EngineeringIBM, Google, and AWS shipped fault-tolerant quantum prototypes this quarter. What it actually means for cryptography, optimisation, and your product roadmap.
Read
EngineeringNext 16, React 19.2, and the Turbopack production builds we've been promised. Three things that quietly changed how we ship product code.
Read
CloudThe component model went 1.0 in February. Suddenly you can write a function in Rust, ship it to Fastly, Cloudflare, AWS, and your laptop — without rewriting.
Read
AILlama 4, Mistral Large 3, and DeepSeek-V4 are within striking distance of the frontier on most benchmarks. What it means for self-hosted AI in production.
Read
EngineeringML-KEM and ML-DSA are now the NIST defaults. Browser support is at 92%. If your TLS, VPN, or signing chain is still RSA-only, you're already behind.
Read
EngineeringCursor, Copilot Workspace, and Claude Code rewrote our hiring and onboarding loops in 2026. Here's what the new senior engineer actually does all day.
Read
StrategyFigure 03 and Tesla Optimus Gen 3 shipped to commercial pilots this year. After visiting four deployments, here's the honest gap between the demo and the floor.
Read
AIMost LLM features collapse in week three because no one defined 'good' in measurable terms. Here's the eval-first workflow we run on every AI build.
Read
EngineeringWhat broke, what didn't, what we'd do differently. Turbopack, async params, and the new image pipeline — from the trenches.
Read
DesignWhy we generate tokens from a single source-of-truth and treat the design system as a published package — not a Figma file.
Read
CloudRight-sizing, reserved capacity, and the three architectural choices that quietly double your cluster spend.
Read
MobileConflict resolution, sync queues, and the data model that makes spotty 4G a non-event.
Read
StrategyWhy we replaced the 60-slide deliverable with a one-page plan and a one-hour walkthrough — and clients sign faster.
ReadThe best of what our team writes, in your inbox. Unsubscribe in one click.