AI Needs Radar
Weekly automated analysis of global tech community discussions, uncovering real user needs gaps to help product people spot AI tool opportunities.
Top 10 Needs by Heat
Overall tab cross-cuts all categories; each category tab is ranked independently
Using Claude Code, Codex, and Aider together causes conflicts in the same project.
Agents execute commands, send emails, and modify databases solely based on prompt constraints.
Agents get stuck in error loops, wasting tokens and time until manually killed.
Skills reload every session, wasting context without knowing which are unused.
Agent grep reads entire code, exploding tokens and pulling in irrelevant files.
Thousands of hours of video recorded, but finding a specific scene relies on memory.
Team members ask humans directly instead of querying AI first.
AI writes an ops panel easily but might directly operate the production database.
Each user command reloads the same skills, causing insufficient context.
Want to advance education but school info is scattered and messy.
Deep Dive by Category
Fine-grained mining across seven categories — pain points, gaps in existing solutions, product opportunities, and competitive landscape
Colleagues waste time with AI-answerable questions
↑ risingTeam members ask humans before AI for trivial questions
Auto-generate step-by-step docs from screen recordings
↑ risingRecorded videos still need manual screenshots and SOP writing
Notes scattered across four apps, impossible to find
↑ risingNotion, Obsidian, WeChat favorites scattered, can't find half-year-old ideas
Distracted by editor while trying to focus on writing
→ steadyEvery computer session becomes fixing blog toolchain instead of writing
Opportunity Radar (Solo & Small Team)
A 2×2 grid of demand strength vs. competition density. Solo and small-team share the same lens with different go-zones. Big-company view is below.
- ▸Can't see which skills and tokens agent uses
- ▸Code search grep is too primitive for agents
- ▸AI agent reloads skills wasting context
- ▸Fragmented lock file support across package managers
- ▸Wi-Fi router reconnection triggers light strip
- ▸Colleagues waste time with AI-answerable questions
- ▸Auto-generate step-by-step docs from screen recordings
- ▸Notes scattered across four apps, impossible to find
- ▸Unsure which grad school to choose
- ▸No immersive practice for new language
- ▸Video training for volunteers but VPS can't handle
- ▸Content scraped by AI, no traffic back
- ▸Insurance AI gives product names but no selection help
- ▸AI teaches outdated Rust syntax
🏢 Big Company View (Enterprise)
Directions that need strategic synergy, scale, and brand — beyond what solo or small teams can pursue
Multiple AI agents can't collaborate
Using Claude Code, Codex, Aider simultaneously, they overwrite each other in the same project
No security boundary for sensitive agent actions
Agents can execute commands, send emails, modify DBs, relying solely on prompt constraints
No reliable fallback when AI agent errors
Agent loops on error branches, wasting tokens and time
AI-generated internal tools are low quality and dangerous
Writing an ops dashboard with AI is easy, but it might directly operate production databases
Smart home switch status falsely shows success
Mi Home shows light off, but it's still on
Apple Watch battery can't last long hike
Sports mode drops below 20% in 6 hours, needs mid-trip charge
What's New This Period
Just-surfacing needs and accelerating trends
- ▸Zero-based men's skincare beginners can't find first step
- ▸AI teaches outdated Rust syntax
- ▸Multiple AI agents can't collaborate
- ▸No security boundary for sensitive agent actions
- ▸No reliable fallback when AI agent errors
- ▸Can't see which skills and tokens agent uses
- ▸Code search grep is too primitive for agents
- ▸Year of video footage lost on hard drive
- ▸Colleagues waste time with AI-answerable questions
- ▸AI-generated internal tools are low quality and dangerous
- ▸AI agent reloads skills wasting context
- ▸Unsure which grad school to choose
- ▸Auto-generate step-by-step docs from screen recordings
- ▸Notes scattered across four apps, impossible to find
- ▸Insurance AI gives product names but no selection help
- ▸Apple Watch battery can't last long hike
- ▸Video training for volunteers but VPS can't handle
- ▸Content scraped by AI, no traffic back
- ▸Cross-platform CLI lacks unified DDC command support
- ▸Smart home switch status falsely shows success
Selected User Voices
Real user comments from global tech community discussions, translated and curated
Using Claude Code, Codex, and Aider together—they each do their own thing, overwriting each other in the same project. I edit one file and all three agents jump in. So annoying.
Agents can execute commands, send emails, modify databases—all relying on prompt constraints. If the prompt gets bypassed, company data is toast.
Agents hit a wrong branch and retry in a loop, wasting tons of tokens and time. In the end, I have to kill the process manually.
Skills reload every session, context is never enough, and I have no idea which skills are unused—it's a total black box.
Agent grep reads entire code, token usage explodes, and it often pulls in unrelated files. Searching for a function name can bring up the whole project.
Recorded thousands of hours of video, but finding a specific scene relies on memory. Tons of footage on the hard drive, but I can never find what I need.
Someone in the team asks me directly before even trying the AI. I've become a human search engine.
Using AI to write an ops panel is great, but it might directly operate the production database—scares me to death.
Past Reports
Browse past editions to compare how needs evolve
Subscribe to the Weekly
Weekly report delivered every Monday via Substack. No spam, one-click unsubscribe.