This Week’s Questions

  • In 2026, am I directing my greatest energy toward building core competencies in Agent + AI search + automation? What specific actions and strategies am I taking right now?

  • If I could keep only one project alive, which would it be—and why?

  • Among what I’m doing today, is there one thing that—were it to succeed—would fundamentally reshape the entire landscape, not just add another revenue stream?

  • If I rebuilt my current team from scratch and kept only one person, who would it be—and why?

  • Did I generate linear returns this week—or nonlinear returns?

  • What’s the single most important thing I learned this week that will matter more over the next 12 months than it does today?

  • Which tasks or projects, if shut down right now, would make me breathe easier?

The Agent Era

At the AGI-Next closed-door summit, top AI researchers dove deep into scaling law evolution, reinforcement learning applications, and the paradigm shift from conversational bots to autonomous agents.

Last year, I spent many hours with Xiang Yang exploring agents. We converged on one clear insight: Whoever deploys thousands of agents—each autonomously completing specialized tasks daily—will command vastly greater productivity.

Experts predict that by 2026, agents will reliably automate workloads equivalent to 1–2 weeks of human effort—making this the pivotal year when agents begin delivering massive economic value. They’ll no longer just assist; they’ll act like elite individual contributors.

AI is becoming a hirable worker.

At the summit, Tang Jie, Yang Zhilin, Lin Junyang, and Yao Shunyu jointly concluded: When Reasoning + Agentic behavior + Coding are fused at the system level, AI transcends “tool” status and becomes an executive agent.

The dominant trend emerging in 2026 is a full paradigm shift—from Chat to Doing. The goal is no longer just answering questions, but enabling anyone to use AI to complete a concrete, complex task—not just retrieve information.

Such agents meet the functional definition of AGI: true general-purpose intelligences. Specifically, they exhibit:

  1. Spatiotemporally consistent multimodal understanding
  2. Controllable online learning and adaptation
  3. Verifiable reasoning and long-horizon planning & execution
  4. Calibratable reflection and metacognition
  5. Strong cross-task generalization

How should individuals embrace this?

  1. Upgrade your mental model: Recognize AI’s evolution from “chatbot” to “task executor.” Start delegating lengthy, repetitive tasks—like weekly reporting or content workflows—to AI (e.g., via Claude Code’s skills).
  2. Treat AI tool fluency as non-negotiable: As Yao Shunyu put it, those who wield AI tools will replace those who don’t—just as programmers replaced slide-rule users. Make it habitual.
  3. Master context provision: AI’s value scales dramatically with the quality of input you give it. Learn to feed it your chat history, preferences, and environmental signals—so it knows you, and delivers truly personalized output.
  4. Stay patient and persistent: Tang Jie’s “coffee spirit” applies here too. AGI is a decades-long race. Ignore short-term hype cycles. Invest steadily. Dare to explore unconventional, non-consensus ideas.

A New Hiring Framework

We’ve refined our hiring criteria and methodology to reflect the Agent Era.

Who we’re looking for:

  1. Real, demonstrable AI mastery—not just familiarity, but results
  2. Ability to translate new tools into measurable efficiency gains or structural business change
  3. Proven capacity for sustained learning and methodological rigor—not one-off luck

We seek colleagues who treat AI as a productivity lever—and proactively redesign their own workflows around it.

Interview Standard One: AI Competence (Weight: 50%+)

1. Evidence of standout AI application

Ask directly:

  • “Tell me about a real case where you used AI to significantly improve your work’s speed, quality, or impact.”
  • Bonus if it was self-initiated—not mandated by your company.

What matters isn’t which model you used or how fancy your prompt was—it’s whether AI fundamentally changed how you worked.

2. Has AI become a tool, not just a chatbox?

Look for evidence of reusable, persistent AI-powered systems:

  • An automated data-cleaning + weekly report generator
  • A semi-autonomous content creation → fact-checking → publishing pipeline
  • An AI assistant that handles 30–50% of your recurring manual labor

This is the critical threshold:

  • Baseline: AI = conversational interface
  • Advanced: AI = embedded component of your production system

3. Clarity of underlying thinking

Probe deeply on three points:

  1. What pain point drove you to build this?
  2. How did you decompose the problem—and decide what part AI should own?
  3. What was the outcome? Did you iterate?

From these answers, we assess: depth of AI understanding, logical structuring ability, and realistic awareness of AI’s limits.

Why prioritize AI competence? Three reasons:

  1. Reveals attitude: fear, performative trend-chasing, or genuine view of AI as tameable infrastructure
  2. Signals learning agility: self-driven adoption and weaponization of new tools
  3. Tests boundary awareness: knowing when to deploy AI—and when humans must step in

Ultimately, we look for candidates who’ve replaced ≥50% of their own repetitive work with AI—not once, but sustainably.

Interview Standard Two: Operational Strength (Pick one)

No need for universal excellence—just one clearly demonstrated strength, backed by a vivid case:

  1. Operations (content, growth, users, channels)
    • Clear objective-setting
    • Metric decomposition skill
    • Systematic optimization experience
  2. Project Management
    • Ability to coordinate cross-functional teams
    • Comfort navigating ambiguity and driving progress
    • Experience bridging role-specific communication gaps
  3. Data Analysis
    • Data-informed decision-making (not just reporting)
    • Full feedback loop: analysis → action → validation
    • Capacity to distill complexity into clear, data-grounded narratives

One requirement: This strength must be showcased with a real, tangible case—as compelling as any AI story.

Final Screening Principle (Critical)

Does the candidate rapidly adopt and operationalize novelty?

If someone:

  • Actively restructures their workflow around AI
  • Uses it in ways distinct from “typical users”
  • Can articulate their reasoning transparently

…then their learning velocity, adaptability, and long-term ceiling are almost certainly high.

Start interviews with this lean, direct question:

“Tell me about a time you used AI to meaningfully change how you work.
How did you do it before AI? What changed most after?
Ideally, it’s something you built—a tool or process.
Most importantly: What were you thinking—and what actually shifted?”

What Makes a Great Salesperson

The hardest part of selling has never been polishing your pitch or speaking fluently.

It’s detecting the customer’s unspoken, deeper needs—and responding with flexible, authentic resonance.

This gap widens dramatically with standardized products.

Three exceptionally difficult, defining challenges:

1. Rapidly uncovering the real need

Customers voice surface-level concerns: price, features, competitors, process. But those rarely drive decisions.

What actually moves them? Risk aversion, accountability pressure, anxiety about outcomes, fear of failure—or hunger for success.

Great sellers spot these in just a few exchanges. That’s the first inflection point.

2. Delivering personalization within standard constraints

The product doesn’t change. Its features don’t change. But every buyer’s context, priorities, and psychology do.

Top sellers make customers feel the solution was built for them—without altering the product. Weak sellers recite specs robotically, delivering identical monologues to everyone.

3. Making the customer feel ownership—not persuasion

This is the hardest: creating a dynamic where the buyer feels, “This is what I need,” not “You made me need this.”

When buyers sense they’re being led, sold, or pushed, resistance and suspicion linger. When they believe they arrived at the decision themselves, commitment solidifies—and deals close cleanly.

In recent sales engagements, the contrast was stark:

  • Average sellers launch into “output mode”: relentless feature-dumping, case-study reciting, and self-promotion. In today’s overloaded, skeptical world, this often backfires.
  • Exceptional sellers say little—but listen deeply. They’re the ones who know what the customer is thinking, fearing, and wanting—before the customer says it aloud.

Ten Ways Startup Death Happens

  1. Founder distraction: Too many initiatives → organizational entropy
  2. Founder stagnation: Cognitive lag behind tech and market shifts
  3. Cash collapse: No viable revenue or funding path
  4. Team decay: Erosion of expertise and execution capability
  5. Strategic drift: Constant pivots, no north star
  6. False innovation: Heavy investment in fake problems or pseudo-solutions
  7. Bloat: Too many managers, too few doers
  8. Financial mismanagement: Uncontrolled spending, poor cost structure
  9. Broken business model: No path to sustainable profit
  10. Core team fracture: Co-founder or key talent implosion

Strategy Must Be Steady—Tactics Must Be Fast

In startups, clarity on direction, model, and north star metric is non-negotiable—and shouldn’t shift lightly.

Within that stable frame, tactics thrive on speed:

1. Anchor the unchanging

  • Define vision and core objectives explicitly
  • Use them to set firm boundaries—clarity here prevents drift

2. Decide what you will and won’t do

  • “Yes” means focus. “No” means protection.
  • Conserve bandwidth. Resist scope creep.

3. Iterate relentlessly within the frame

  • Test fast. Fail small. Learn faster.
  • Validate assumptions with micro-experiments—not grand theories.
  • Let small wins inform bigger bets.

Six AI Mindsets for 2026

  1. Use the best models
    Yes, costs may rise—but so do possibilities. The real ROI isn’t lower bills—it’s your saved time and mental energy. Try Claude 4.5, Gemini 3, or GPT-5.2.

  2. Adopt and master elite tools
    Don’t chase every new app. Pick 3–4 professional-grade tools (e.g., Claude Code, VS Code, Warp, Raycast, Cursor) and go deep. Build durable habits. Stay curious about others—but anchor in excellence.

  3. Bring AI into all thinking—not just “big tasks”
    Use top-tier models for daily reflection, conversation, judgment calls—even casual planning. Your everyday AI dialogue quality shapes your cognition.

  4. AI First
    Before writing anything—idea, spec, strategy, email—co-create with AI first. Then refine with your judgment.

And dare to build with AI: landing pages, internal tools, browser extensions, apps. I built both an AI browser and mobile app—both shipped fast and worked well.

This isn’t about offloading work. It’s about reconstructing your capability stack—and upgrading your productivity architecture.

  1. Ignore the “dumb outputs”—focus on your judgment
    Wrong code? Low-quality drafts? Off-base opinions? All normal. Like working with a brilliant but fallible human, AI’s value lies not in perfection—but in your ability to evaluate, correct, and steer.

  2. Cultivate aesthetic discernment
    True taste takes time—but accelerates with deliberate practice:

    • Study masters in your field—deeply, repeatedly
    • Analyze world-class work in your domain: deconstruct it, reverse-engineer its logic
    • Refine your eye for quality: Is this website elegant or cluttered? Is this message dense or flabby? Does this solution feel restrained—or cheap?

Here, “aesthetic” isn’t about art alone. It’s your calibrated sense of what’s good, effective, and elevated—across interfaces, language, logic, and execution.