Apurva Shukla/writing

11-15-25: Internet goodies I read last week

  1. How to Market Yourself

    • Marketing yourself isn’t about becoming a celebrity. It’s about making it easy for the right people to recognize and remember your value. Build a personal brand (identity + opinions) and be relentlessly consistent across photo, name/handle, and keywords. Own a domain twice: yourname.com to control your narrative, and a focus area to “plant your flag.” Show up repeatedly so you enter others’ consideration set; you can always pivot later. Market two things: your business value (keep a brag doc, speak WIIFT) and your coding skills (do cool work, cover the basics). Portfolios are optional—proof of work can be OSS, focused blogging, hackathons, or deep projects. In public: pick 1–2 channels, tell the truth, protect confidences, tailor messages, and shift from outbound to inbound. At work: log metrics, deliver standout updates/demos, and launch signature initiatives—while sharing credit. Ignore vanity; prioritize relationships.

  2. notes from Naval (2025)

    • Key takeaways from a targeted read of Naval’s Almanack. Wealth: prioritize wealth over money or status; get paid for judgment by building equity and outsourcing everything below a high hourly rate; climb leverage ladders—from labor to capital to code—to scale what society wants. The Money:Wealth exchange rate is often far higher than people think, though present cash still matters. Critiques: reading only “for the present” misses full arguments, and Naval’s judgment advice feels aphoristic over applied. Happiness: self-care order—physical, mental, spiritual, family, then world; pair exercise with meditation (walking works); use discomfort training (ice baths) to build mental strength. Reading queue: Thinking Physics, Six Easy Pieces, The Sovereign Individual; finally Meditations and Poor Charlie’s Almanack; plus Ted Chiang and Snow Crash. Quote: "I don’t care if the world is imploding and melting down, it can wait another thirty minutes until I’m done working out."

  3. My 2025 New Mac Setup

    • A fresh 2024 MacBook Pro setup distilled. Core idea: make Raycast the “super app,” replacing small utilities (Caffeine, window management) with extensions. Lean into AI: Claude Desktop plus iMCP to read iMessages, Windsurf for AI-first coding, and lightweight voice dictation tools. System bootstrap: use Arc as default browser and trigger screen-share to pre-grant permissions; pick Warp or Ghostty for terminals; install ZSH, Homebrew, and Apple CLT early; batch-install via brew. Sensible macOS/Finder/keyboard/dock tweaks reduce friction. Dev stack: GitHub CLI with SSH, Python via Astral’s uv (skip pyenv/Anaconda), Node via fnm and pnpm, z for jump navigation, powerlevel10k, and nicer diffs with diff-so-fancy/delta. Apps: Podman/Colima instead of Docker Desktop; noTunes + VLC/Audacity; Shottr for screenshots/OCR; Pure Paste; Alfred clipboard; Rectangle/Loop; Amphetamine; Zoom/Loom; Superhuman; Obsidian/Notion; Cursor editor with sync. Includes links to many alternative setup guides.

  4. Nano Banana can be prompt engineered for extremely nuanced AI image generation

    • Google’s “Nano Banana” (Gemini 2.5 Flash Image) is an autoregressive image model that excels at following complex prompts. In tests, it nailed multi-edit image updates, maintained subject identity from reference photos (e.g., Ugly Sonic shaking hands with Obama), and even improved composition when prompted with newsroom-style cues. Its 32k-token context and Gemini 2.5 Flash text encoder let it parse long, structured inputs—Markdown rules, HTML, and detailed JSON character specs—surprisingly well. It can generate images programmatically via API (~$0.04 per 1MP) or free in Gemini (with watermark); a lightweight gemimg Python wrapper simplifies usage. Weak spots: mediocre style transfer (e.g., “make me into Studio Ghibli”), lenient IP/NSFW moderation, and occasional text/code accuracy quirks. Notably, prompt-injection experiments suggest a Markdown-like system prompt and guardrails shaping outputs. “It’s not silly if it works” neatly captures the model’s engineer-friendly prompting ethos.

  5. The Two Jobs of a CPO

    • Being a Head of Product means doing two hard jobs at once: build a high-quality product culture and align tightly with the CEO on product strategy. Culture—processes, rituals, hiring, templates—can be improved iteratively; strategy must be right from day one, or you’ll be replaced. Delegate operational muscle to seasoned product builders (not generic ops), because nuances vary by business. For strategy, be decisive, hire and train people who “get the plan,” and expect them to seize opportunities without hand-holding. Strategic alignment doesn’t mean parroting the CEO; it means incorporating their ideas and elevating the team’s best ones. This dichotomy explains why CPO roles are hard to fill: companies hire culture-builders but fire for misalignment. Communicate constantly with your CEO to keep the department pointed in the same direction. Quotes: "One of the most important ways to run strategy is to just be a dictator."

  6. Fear of Sales

    • Bottoms-up adoption feels cool, but enterprise sales is often what scales SaaS: DataDog drives 72% of ARR from large enterprises. Technical founders fear sales due to unfamiliarity, the myth that it’s art not science, and cultural worries. In reality, modern sales is a repeatable, metric-driven system; a strong leader brings playbooks, pipeline rigor, hiring and comp models. Start small: hire a sales leader, add 1–2 reps, measure payback and conversion, and use sales to feed product/competitive insights. Biggest pitfall: overhiring before product-market fit or a repeatable model, which torches cash. The core choice: if you want to win your market and penetrate big companies, begin building sales as soon as you see PMF—don’t wait. You may cycle through 1–2 leaders; that’s okay. Quote: "Sales is a giant process engineering exercise which has benefited from 40 years of iteration."

  7. What is Amazon?

    • To grasp Amazon, start with Walmart: a bounded search optimizing finite shelf space for selection, price, and inventory. Amazon removed that constraint, running an unbounded search by aggregating SKUs and sellers (Marketplace), then turning every bottleneck into a platform—AWS, Catalog API, FBA—per Bezos’s mandate that teams expose services via interfaces. This scaled fast and checked bureaucracy but invited gaming and a major misstep: ads. Sponsored listings push high‑margin, ad‑savvy products over the best ones, degrading customer outcomes. With infinite sellers, quality control is a constant war. Modular by design, Amazon shrugs at antitrust; its future hinges on aligning seller incentives with its customer obsession. Quote: "In the world of infinite shelf space, the ranking algorithm is practically the entire merchandising strategy."