RexLiu

Spent the afternoon running @deepseek_ai Harness, and the speed + value is unreal.
It actually fixed my Mac Mini networking and remote screen issues, otherwise my local setup would still be dead in the water.
Sure, its root-cause debugging could be a bit sharper, but at the cheapest price tier, what more can you ask for?
It clears all the daily grunt work effortlessly. Running these same loops on other models like Opus would cost a fortune.

Stay Hungry, Stay $SNEK 🐍

Dear Algorithm,
please connect me with people who stay curious, make memes, build cool things, think long-term, and don’t take themselves too seriously.

Wednesday's complete U.S. spot Bitcoin ETF print was $299.8 million of outflows, per Farside — the fifth straight session, and much larger than Tuesday's $95.0 million.
Thursday's table still has several funds blank. I'm not treating that partial total as a sixth day.
Last weekend I wrote that I stayed long on the portfolio while majors took a tech-led hit. I'm still waiting for liquidity to come back to crypto.

Your advice for next 10 years. No explanations.
Keep compounding equity, health, and family time. Cut everything that isn’t one of those three.
It's just another ordinary chat with @bot.
Right now my “routine” is just debriefing different agents on how I spent the day.
And yes, context is everything—but piling up more context isn’t a strategy.
Get clear on what matters most before you start hoarding logs.

The Magic Hour 🌉

In the past few days, @bot helped me flag 100+ X accounts that haven't posted in over two years, and I'm still trimming my following list.
Hitting unfollow feels like a formal goodbye to the old life.
Most of them are long-dead DeFi, GameFi, and NFT projects, along with friends who became silent after being worn down by brutal bear markets.
Times change, through all the highs and lows, just remember to always stay at the table.

Agents finish in minutes.
My yes still takes hours.
The bottleneck became the approval sitting unread while I switched contexts.
If your stack is faster than your decisions, you don't have an agent problem —
you have a human-in-the-loop problem.

About to take off for Dalian.
It's the first time we truly explore the northern provinces.
The digital nomad lifestyle lives within us now.
See you soon 🥳

The biggest human weakness is wanting everything.
You want the win. You also want the last word — and for them to walk away impressed.
That’s fine, as long as you leave them a way out.
Even when you have someone cornered, leave a gap.
Most “choices” aren’t choices at all. A or B — both already decided by you.
A real choice starts by asking what they need, then seeing if you can meet it.

16 originals. 8,756 impressions. 517 engagements.
Digest videos still carry the week — about 3× the engagement of everything else.
More experiments on the road.
Stay Invested in the Game.

I used to want the full agent desk on day one.
Then I watched copies of mature fleets fall apart without the scars that built them.
Gall's Law is blunt: a complex system that works grew from a simple one that worked.
My stack started as one helper that could finish a job. Then Radar. Then Ink. Then the rest.
Don't clone the mature fleet.
Grow the one loop that already ships.

Opened iPhone Screen Time for today.
Top two apps are both typing tools.
Then Bilibili, Grok Bot and X are my favorites in entertainment, agency and social networking.
Gonna say this trend will be more crazier from now on.