Tahosin.
the library

20 pieces, hand-picked.

Only the ones where I actually shipped a project or solved a real problem. News and aggregators are filtered out. Tap any to read.

  1. TanStack Query v5: Why status === 'pending' Broke Your Loading States (and the 3 Patterns That Fix It)

    v5 renamed `loading` to `pending` and that one rename quietly broke a bunch of UIs. Here are the three patterns I switched to — `isFetching`-aware skeletons, suspense boundaries, and a tiny custom `useQueryUI` hook.

    dev.to
  2. Streaming Speech-to-Text with OpenAI in 2026: Moving Beyond Whisper

    Whisper is no longer the best STT primitive in 2026. Here's how I migrated to OpenAI's new streaming STT, the latency budget I worked with, and the chunked-mic-input pattern that finally felt real-time.

    dev.to
  3. Next.js 16: Revalidating Per-User Dynamic Fetches on Demand (3 Patterns That Actually Work)

    Three patterns I actually use to revalidate per-user dynamic fetches in Next.js 16 — tag-based, on-demand, and segment-level. With code, edge cases, and the gotchas I hit in production.

    dev.to
  4. TanStack Query v5: Why `status === 'pending'` Broke Your Loading States (and the 3 Patterns That Fix It)

    v5 renamed `loading` to `pending` and that one rename quietly broke a bunch of UIs. Here are the three patterns I switched to — `isFetching`-aware skeletons, suspense boundaries, and a tiny custom `useQueryUI` hook.

    hashnode
  5. Streaming Speech-to-Text with OpenAI in 2026: Moving Beyond Whisper

    Whisper is no longer the best STT primitive in 2026. Here's how I migrated to OpenAI's new streaming STT, the latency budget I worked with, and the chunked-mic-input pattern that finally felt real-time.

    hashnode
  6. Next.js 16: Revalidating Per-User Dynamic Fetches on Demand (3 Patterns That Actually Work)

    Three patterns I actually use to revalidate per-user dynamic fetches in Next.js 16 — tag-based, on-demand, and segment-level. With code, edge cases, and the gotchas I hit in production.

    hashnode
  7. I Ran 1,000+ Real Tests on Gemini 3 Flash (Preview) vs Gemini 2.5 Pro — Flash Won

    I ran 1,000+ real-world prompts across Gemini 3 Flash (Preview) and Gemini 2.5 Pro — coding, RAG, structured output, vision, agentic loops. Flash won on speed-per-dollar in every category that mattered for indie infra.

    dev.to
  8. $0/Month, 5 AI Apps, All on Gemini: Here's Exactly What the Free Tier Gives You (and What Breaks First)

    I shipped 5 AI apps on Gemini's free tier. Here's the exact architecture, cost breakdown, and where it actually breaks in production.

    dev.to
  9. HOCKS AI: I Open-Sourced a Full AI Platform With Chat, Vision, Video Analysis & Website Generation — Runs at $0/Month

    HOCKS AI is a free, open-source multi-modal AI platform built with React, Firebase, Google Gemini, and OpenRouter. It features streaming chat, image analysis, video understanding, website generation, and a per-user memory system — all running at nearly zero cost. Here's the full technical breakdown.

    dev.to
  10. Building a Free AI Chat Platform: The Architecture Behind HOCKS AI

    HOCKS AI — the architecture diagram. Streaming chat, image analysis, video understanding, website generation, per-user memory. All on a $0/mo budget. Here's exactly how the pieces fit.

    dev.to
  11. I Built a Free Multi-Modal AI Platform with OpenRouter, Gemini & Firebase — Here's How

    Build a free multi-modal AI platform using OpenRouter free models for chat, Gemini for vision, and Firebase for hosting. Full architecture guide with streaming chat, memory system, and admin dashboard.

    dev.to
  12. EcoSense AI: Know Your Carbon Footprint in 60 Seconds

    For the DEV Weekend Challenge I built EcoSense AI — an AI-powered carbon footprint analyzer that gives you a personalized environmental impact analysis in 60 seconds. Architecture, prompt design, and how I kept it free.

    hashnode
  13. I Built a Free Multi-Modal AI Platform with OpenRouter, Gemini & Firebase

    HOCKS AI handles chat, image analysis, video analysis, and full website generation — without paying a fortune for API costs. Production-ready architecture, the prompt patterns, and the cost-per-call numbers.

    hashnode
  14. How I Built an AI Carbon Footprint Analyzer for Earth Day 2026

    Building EcoSense AI: the prompt-engineering, the Backboard memory layer, the Solana verifiable-claim pattern, and the deploy story on Cloudflare Pages.

    dev.to
  15. Build a Free AI Writing Tool with Next.js and Google Gemini (2026 Guide)

    Step-by-step: build a free AI writing tool using Next.js 16 and Google Gemini API. Complete code, server-side proxy pattern, and the rate-limit handling I actually use.

    dev.to
  16. Why OpenClaw Skills Are the Most Underrated Feature in Personal AI

    OpenClaw Writing Challenge entry. Most personal AI is locked behind chat windows. Skills change that — they let an assistant carry one focused superpower into every conversation. Here's why I think they matter, drawn from building EcoBot.

    dev.to
  17. EcoBot: A Personal Carbon Footprint Tracker Built with OpenClaw

    OpenClaw Challenge submission. EcoBot is a personal carbon-footprint skill that gives you a one-line nudge on every action you log. Skill spec, test prompts, and what surprised me about OpenClaw's runtime.

    dev.to
  18. EcoSense AI: Know Your Carbon Footprint in 60 Seconds

    DEV Weekend Challenge: Earth Day Edition. EcoSense AI is a personal carbon analyzer powered by Google Gemini, with persistent memory via Backboard and verifiable claims via Solana.

    dev.to
  19. PromptCraft AI: Free Prompt Generator for Midjourney, DALL-E 3 & Stable Diffusion

    PromptCraft AI: structured prompts with style presets, aspect ratios, camera angles, and negative-prompt sanitization. Zero signup. Here's the prompt-engineering pattern under the hood.

    dev.to
  20. I Built 6 Free AI Writing Tools in One Weekend - No Signup Required

    Resume builder, cover letter generator, email writer, content rewriter, summarizer, tone-shifter — all free, no signup. Here's the shared scaffold and the prompt patterns I reused across all six.

    dev.to