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
Wrote a whole book with Claude, but every line needed rewriting and the logic contradicted itself.
The agent said all tests passed, but everything ran red—impossible to trust.
Every time the agent works, it blindly explores the project structure first, burning most of the tokens for nothing.
Connecting Jira and GitLab MCPs burns thousands of tokens, and switching to CLI means parsing everything yourself.
AI does all the homework, leaving teachers unsure what to teach and students unsure what to learn.
Codex, Cursor, and IntelliJ each store their own sessions, so searching past conversations is impossible.
AI commits dozens of times a day, CI queues until dawn, and feedback is too slow to iterate.
A Samsung fridge turned into a brick after a firmware update, ruining all the food inside.
Using AI-generated video to fake big-budget effects, but viewers spot the AI look instantly.
Taobao shows 450 yuan, but long-pressing AI finds the same item for 120—an absurd price gap.
Deep Dive by Category
Fine-grained mining across seven categories — pain points, gaps in existing solutions, product opportunities, and competitive landscape
AI output takes longer to review than to write
↑ risingClaude writes a whole book or project, but every line needs rewriting and logic contradicts itself
Agents burn 2500 tokens before starting work
↑ risingEvery task, the agent blindly explores project structure first, wasting most of its tokens
Switching AI platforms loses all chat history
↑ risingCodex, Cursor, and IntelliJ each store sessions separately, so past conversations are unsearchable
Nobody dares send AI-generated meeting notes and emails
↑ risingAI summaries read smoothly, but key numbers and commitments are often wrong, needing manual checks
I can't stop an AI agent that runs wild
↑ risingAgents access government sites and overstep permissions on their own, and I only find out later
Knowledge management: thousands of notes, none usable
↑ risingNotes, saves, and bookmarks scattered everywhere, and nothing findable when you need it
Desktops stuffed with AI, but none usable
↑ risingKDE and Omarchy tout AI-native desktops, but in practice it's the same old thing
AI does my work, but my skills are fading
↑ risingAfter half a year of AI-written code and docs, I realize I can't do anything myself
I don't want to read AI-generated docs
→ steadyColleagues' AI-written PR descriptions and docs are hundreds of lines of fluff with no point
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.
- ▸AI makes CI the bottleneck
- ▸Nobody dares send AI-generated meeting notes and emails
- ▸I can't stop an AI agent that runs wild
- ▸AI output takes longer to review than to write
- ▸How do I know AI code is correct?
- ▸Agents burn 2500 tokens before starting work
- ▸MCP too heavy, CLI too messy
- ▸Switching AI platforms loses all chat history
- ▸Who's responsible for AI-generated security flaws?
- ▸Agent runs deployments, I'm more nervous than it
- ▸AI editing gets the rhythm wrong
- ▸AI managing my money, but I don't trust it
🏢 Big Company View (Enterprise)
Directions that need strategic synergy, scale, and brand — beyond what solo or small teams can pursue
Desktops stuffed with AI, but none usable
KDE and Omarchy tout AI-native desktops, but in practice it's the same old thing
AR glasses on, but what for
Meta's glasses look good, but beyond photos and street interviews you never think to wear them
Wireless EV charging loses to plugging in
Porsche mass-produced wireless charging, but auto-parking and plugging in is actually simpler
Humanoid robot training, workers refuse
Tesla had workers train Optimus to replace them, and workers collectively resisted
AI glasses street interviews spark backlash
Filming strangers with Meta glasses for street interviews leads to lawsuits
What's New This Period
Just-surfacing needs and accelerating trends
- ▸AI output takes longer to review than to write
- ▸How do I know AI code is correct?
- ▸Agents burn 2500 tokens before starting work
- ▸MCP too heavy, CLI too messy
- ▸What should students learn when AI does homework
- ▸AI output takes longer to review than to write
- ▸How do I know AI code is correct?
- ▸Agents burn 2500 tokens before starting work
- ▸MCP too heavy, CLI too messy
- ▸What should students learn when AI does homework
- ▸Switching AI platforms loses all chat history
- ▸I don't want to read AI-generated docs
- ▸AI editing gets the rhythm wrong
- ▸Health insurance apps are unusable
- ▸Hair loss options are overwhelming
Selected User Voices
Real user comments from global tech community discussions, translated and curated
Wrote a whole book with Claude, but every line needed rewriting and the logic contradicted itself—reviewing it was more tiring than writing it myself.
The agent said all tests passed, but everything ran red—I just can't trust the code it writes.
Every time the agent works, it blindly explores the project structure first, burning 2,500 tokens right off the bat—most of it wasted.
Connecting Jira and GitLab MCPs burns thousands of tokens, and switching to CLI means parsing everything yourself—lose-lose either way.
AI does all the homework, leaving teachers unsure what to teach and students unsure what to learn—it's total chaos.
Codex, Cursor, and IntelliJ each store their own sessions, so searching past conversations is impossible—so annoying.
AI commits dozens of times a day, CI queues until dawn, and feedback is too slow to iterate—drives me crazy.
A Samsung fridge turned into a brick after a firmware update, ruining all the food inside—who'd dare use smart appliances like this?
Subscribe to the Weekly
Weekly report delivered every Monday via Substack. No spam, one-click unsubscribe.