Blog

Technical insights, development tips, and thoughts on building AI software

AI Bots Are Repeating the JavaScript Framework Era — Just With Bigger Consequences

Companies are spinning up AI bots like they used to spin up JavaScript frameworks — fast, chaotic, and with zero long-term plan. The result? A growing pile of tech debt disguised as “innovation.” This post breaks down how today’s AI sprawl mirrors the front-end chaos of the 2010s and what smarter builders can do to avoid repeating history.

Read more →

When AI Talks JSON: Why Structured Feedback Finally Feels Real

Getting JSON from an AI used to feel like herding cats. Now with GPT-5, function calling, and strict response formats, it’s finally predictable. Let’s talk about why GPT-4 made it painful, why GPT-5 nailed it, and how Claude Sonnet 4.5 and Gemini 2.5 handle structure today. A year ago, asking an AI for clean JSON was like asking Sam Bankman-Fried for accounting advice — you’d get a confident answer, but you’d still end up in court. That’s changed.

Read more →

How We Built and Launched a SaaS in One Day

Most founders spend months polishing their MVP. I built and launched a full SaaS — frontend, backend, billing, and email campaign — in a single day. Not as a stunt, but as proof that craftsmanship and speed can coexist. Execution isn’t reckless anymore. It’s the new edge.

Read more →

Development Is Now Asynchronous

Coding isn’t synchronous anymore. We don’t write and wait for the compiler — we prompt, iterate, and orchestrate feedback loops across multiple AI threads. The best builders are already running multi-window GPT setups, mapping intent across files instead of pasting code blind. Meanwhile, corporate dev teams stuck behind “no external AI” policies are slowing to a crawl. The future isn’t about typing faster — it’s about closing the loop between idea, prompt, and execution. Development is now async.

Read more →

Building AI Chatbots: Best Practices for 2025

Learn how to build production-ready AI chatbots that actually help your users. From prompt engineering to error handling, we cover the essentials.

Read more →

AI Wrappers Are Still IP (Until AGI Arrives)

AI wrappers aren’t just glue code — they’re where real value, differentiation, and defensibility still live. Until we hit true AGI, implementation matters.

Read more →

Gemini’s Context Window Is Wild — But Soulless

Gemini can juggle more context than any model out there. It’s a technical flex — but man, it still feels like talking to an overachieving intern with no vibe.

Read more →