archive
REX·DAILY
Sunday September 20 Anno 2026 · Ibaraki, Osaka
I Daily Wisdom
Robert Greene The Daily Laws September 20 · Hit from Unexpected Angles

People expect your behavior to conform to known patterns and conventions. Your task as a strategist is to upset their expectations. Surprise them with chaos and unpredictability—which they try desperately to keep at bay. To Sun Tzu and the ancient Chinese, doing something extraordinary had little effect without a setup of something ordinary. You had to mix the two—to fix your opponent’s expectations with some banal, ordinary maneuver, a comfortable pattern that they would then expect you to follow. With the enemy sufficiently mesmerized, you would then hit it with the extraordinary, a show of stunning force from an entirely new angle. Framed by the predictable, the blow would have double the impact.

Daily Law: Operate according to your own rhythms, adapting strategies to your idiosyncrasies, not the other way around. Refusing to follow common patterns will make it hard for people to guess what you’ll do next.

Robert Greene, “What Muhammad Ali Can Teach Us about Success and an Authentic Life,”

The Observer

, July 22, 2015

Steven Pressfield The Daily Pressfield Day 263

MY OWN (SECOND) EPIPHANAL MOMENT I was sitting in my apartment on 15th Street between Seventh and Eighth and I thought, “Instead of hanging myself, why don’t I move to Hollywood?” Until that moment my writer’s dreams had all been about becoming a novelist. No other aim existed for me. It was Hemingway/Kerouac/Henry Miller or nothing. Suddenly I thought, What’s wrong with movies? I love movies. I’ve worked in advertising writing commercials. I know what a camera is. I know what a storyboard is. I know what a script is. I’ve failed as a novelist. Why not head west and fail as a screen writer? Writing Seminar, Nashville, 2019 At once a tremendous weight fell from my shoulders. I looked over at my cat, Mo. He was ready to leave New Yo rk too. Within six weeks we had packed and hit the road.

Leo Tolstoy A Calendar of Wisdom September 20

All good things can be achieved only with effort.

Bad things are easy to do, good things are done only with work and effort.

—DHAMMAPADA, a book of BUDDHIST WISDOM

Enter by the narrow gate; for the gate is wide, and the way is broad that leads to destruction, and many are those who enter by it.

For the gate is small, and the way is narrow that leads to life, and few are those who find it.

—MATTHEW 7:13-14

The way to true knowledge does not go through soft grass covered with flowers.

To find it, a person must climb steep mountains.

—JOHN RUSKIN

Look for the truth; it wants to be found.

—BLAISE PASCAL

A person cries out from pain when he takes up hard physical work after a period of idleness.

Any rest from the struggle for spiritual improvement brings the same pain.

September 2l

The most important and necessary expression of freedom is to give your thought a specific direction.

Work toward the purification of your thoughts.

Without bad thoughts you will be incapable of bad deeds.

—CONFUCIUS

Everything is in heaven’s power, except for our choice of whether to serve God or ourselves.

We cannot prevent birds from flying over our heads, but we can keep them from making nests on top of our heads.

Similarly, bad thoughts sometimes appear in our mind, but we can choose whether we allow them to live there, to create a nest for themselves, and to breed evil deeds.

—MARTIN LUTHER

It is a sin, not only to do bad things, but even to think about bad things.

—ZOROASTER

II The Day Ahead
  • Today · ongoing 研修会(日本語/英語・Sのみ) 2026年間予定
  • Tomorrow · All day Respect for the Aged Day Holidays in Japan
  • Tomorrow · 14:00 個別Mtg 個別Mtg 2026
III Signals & Dispatches

AI / ML

  • @Voxyz_aisource ↗ A practical prompt pattern for Codex and Claude Code keeps HANDOFF.md current with decisions, files, verification, unresolved issues, and next steps.
  • @kobatch_tksource ↗ A U.S. antitrust suit reportedly argues that coordination among OpenAI, Anthropic, and Google on slowing AI safety work could limit competition.

Drug Discovery

  • @igashovsource ↗ LDDM reports a generative structure-based drug-discovery framework validated across five targets, with novel hits confirmed by X-ray.
  • @MariGrinisource ↗ Anthropic is reportedly setting up a biology lab as it expands its AI drug-discovery program.

Biotech

  • @UChannelTVXsource ↗ Anthropic is reportedly pairing Claude with a Bay Area wet lab, robots, and automated equipment for supervised preclinical biology work.
  • @hakataya77source ↗ Tokyo plans up to ¥10 billion over three years to attract five overseas deep-tech firms in quantum, AI, robotics, and biotech; applications close September 30.

Tech

  • @adrgrondinsource ↗ A hands-on report says the iPhone 18 Pro can run a 27B model at roughly twice the speed of the 17 Pro, pointing to a serious on-device AI jump.
  • @signulllsource ↗ Jev-class models are being framed as bounded, parallel, highly compressed systems that could add near-instant adaptive judgment to everyday devices.

Japanese Politics

  • @aneane55source ↗ A Fukuoka mayoral-election debate is urging voters to examine candidate affiliations and policy rather than deciding on youth or novelty alone.
  • @Dnu2sEcF9uLbtJ3source ↗ An Osaka prefectural-politics thread says municipalities outside Osaka City need answers on the agreement scope, investment order, and representation before any same-day vote.
IV Sky over Ibaraki
☁️ +24°C
humidity 76% · wind ↓16 km/h
cloudy morning
V What Rex Can Do Today
  1. Turn the strongest drug-discovery item into a short internal note for the StemRIM team, separating evidence from hype.
  2. Review the upcoming calendar and block a focused 45-minute session for one high-value research task.
  3. Tidy one active project or inbox thread that has been waiting on a small decision.
VI Toward the Long Game
  1. Read the LDDM preprint and write down how its structure-based workflow differs from your current mental model of generative drug design.
  2. Build a tiny agentic literature-review prototype that extracts target, modality, assay, and validation evidence from three papers.
  3. Spend 30 minutes on robot-learning fundamentals: PID, forward/inverse kinematics, and the ACT versus OpenVLA distinction.