barking knee + 10% effort

· Zi Wang · 1 min read

Z / Runner 🏃🏻‍♂️

🏃🏻‍♂️ Kind of an interesting phenomenon this past wk. Knee's barking. Took some photos, uploaded to GPT for a gut check—advice was solid, but reality is a bit of a doldrum. Can't just pretend there's no injury and run through it. less than 3 wks from the half marathon (meant to be a tune-up for the full). Still want to keep some imminent pace, but this weekend was the first time I totally missed the long run. Frustrating. Realized this is the ultimate longevity hiccup. How do you actually practice longevity when the physical hardware glitched?

  • 🏃🏻‍♂️ Don't have the answer right now, i keep running into data scrapping problem as well. are we tracking the right direction? These are hard / hard data crawling problems, but i wonder if there are other grey zone problems that yield more value.

  • 🏃🏻‍♂️ audio transcription is a "solved" problem imo, but video analyzer (per our telegram chat) seems like a new frontier.

  • 🏃🏻‍♂️ oversaturated imo, these are all data brokers and have already "sold" our data to LLM builders.

  • 🏃🏻‍♂️ better training data for LLMs.

  • 🏃🏻‍♂️ $$$ rev potential w/ paying customers.

Stephen / Basketball 🏀

  • 🏀 go to mark (fadil). you need it.

🏀 #better-manus which 10x improvements with 10% effort?

  • crawling against captcha, paywall, auth, proxies, throttles, fingerprints..?
  • 🏀 same concern here. we cannot go too deep on one hard problem (crawling / longevity). keep the eyes on general ai agents – only.
  • better tool use for pdf processing, browser drive, video analyze, audio transcribe, ..?

🏀 #better-manus optimized for these high target sites?

  • business: linkedin, indeed, redfin, crunchbase, nytimes, glassdoor.
  • 🏀 insane how we can't even openly use them (web3!). grey vs magenta zones.
  • social: reddit, instagram, tiktok, redbook, twitter, pinterest.
  • 🏀 low calorie from 1 post, superhuman when aggregated.
  • product: youtube, amazon, trip advisor, maps, yelp, ebay, github.

🏀 google's x-ray search, :site operator, archive.ph caches; request instagram's embed / reddit's .json endpoints; use pymupdf4llm / moviepy / distil-whisper / docking / auto-editor libraries.

  • pay state-of-the-art services: deepgram, capsolver, zenrows, twelve labs, shotstack.
  • pay vertical crawlers: instagram / tiktok (apify); amazon (keepa, rainforest); linkedin (proxycurl, nubela, jobspy); maps (outscrapper); crunchbase (clay).
  • anthropic: research agent, tool search.
  • gemini: deep research agent, computer use, browser use, bash shell.
  • 300ms action model, $25m in 4m.