Tang U-Liang

Steve Jobs' team told him to his face that his ship date was reality distortion. He agreed with their evidence and kept the date anyway. The tape explains a decision rule most founders never learn. December 1985. NeXT is 90 days old, funded with Jobs' own money after he left Apple. At the first company retreat, the team debates slipping the launch from spring 1987 to spring 1988. The pushback is brutal and specific. One team member has receipts: "We've got a person here that said he could do a word processor in six months that's taking three years." Another names the danger: "Reality distortion is reality distortion. It has its motivational value." Build the plan on a fake date, and every design decision made from it gets torn up later. Jobs does not argue the evidence. "Well, George, I can't change the world." He argues something else entirely: "I think we have to drive a stake in the ground somewhere. And I think if we miss this window, then a whole series of events come into play." "We can't sell enough units in 87 to pay for our operating costs." Colleges buy computers in the summer. The campus surveys had already put the ceiling at $3,000. Miss spring 1987, and NeXT sells nothing for a year while burning his money. "We have 18 months. So I don't think we have a company if we don't do this. No matter what I say or anybody else says, that is my deepest belief. If we don't do this, we will not be able to attract great people. We will not be able to retain some of the ones we have." My note: the team argued estimates. Jobs argued conditions. An estimate says when the work might be done, and it invites negotiation. A condition says when the company is dead, and it does not negotiate. That is why the stake held. He anchored the date to the market's calendar and his own runway, not to optimism. The engineers could refute the schedule. Nobody in the room could refute the window. The uncomfortable version for founders: if your deadline comes from your team's estimates, it will move. If it comes from the physics of your market, it was never really a deadline. It is a survival condition wearing one. Steve Jobs at the first NeXT retreat, December 1985. Footage released by the Steve Jobs Archive in May 2026. Founders: if you want your X to do this for your business, check the first reply.

Malaysians who opt out of SOCSO's Lindung 24/7 scheme must sign a liability waiver form, said CEO Azman Aziz. The waiver protects SOCSO and employers from liability for accidents or deaths that occur outside working hours if workers choose to withdraw from the scheme. 🧵1

I gave Fable 5 one job: write custom WebGPU kernels for Gemma 4 inference. It climbed to 84 tok/s, then hit a wall, insisting further optimization was impossible. Hours later, Anthropic rolled back invisible LLM development safeguards, and it hit 255 tok/s. The next day, access to Fable 5 was suspended globally.

The story of my professional life told in 54 seconds No matter how good the pre-planning, no matter how organized it starts out, this is where I end up

David Epstein studied the world's best athletes, scientists, and inventors, and found they all broke the same rule. Here are 10 reasons from "Range" why generalists beat specialists in everything that matters. 1) Specializing late is an advantage, not a delay

What I find fascinating with Claude Fable 5 is it proves once again that large generalist models will outperform vertical ones. On ProofBench (graduate-level formal math benchmark in Lean, where a proof either compiles or it doesn't) Fable 5 beat Harmonic's Aristotle, 77% vs 71%. Aristotle is a system built specifically for formal math + run on its own internal harness, so the generalist beat the specialist on the specialist's home turf. It's the Richard Sutton's "The Bitter Lesson". His whole argument is that across 70 years of machine intelligence research, the methods that win are the general ones that scale with compute. Not the ones where we hand-encode human expertise. Building our own knowledge into the system feels good and helps short term gains but long term it always gets overtaken by bigger model. You can look at Chess, Go, speech, vision, same story every time. First the specialized model wins, then the general one takes over. and btw this is the whole premise of AGI. You don't build one model for math, one for code, one for law. you build a single general model that scales with compute and it learns to do everything

Everyone's banging on about loops When they should be thinking about queues