The PRD is the source code. That is the big shift.
Palo Alto, CA
Joined December 2008
- Tweets406
- Following1.6K
- Followers395
- Likes395
Software moats are done. The buzz the past few days has been around Reverse Engineer Anything (rea.tools/).
What is it? An open source AI agent tool that examines the compiled code of any application and spawns agents to rebuild it or remix it, basically turning Claude Code and Codex into automated clone engines. Feed it compiled code, and agents rebuild the entire application from scratch.
Out in the community people are already remixing games (Minecraft + Attack on Titan) and cloning the entire Adobe Creative Cloud suite (getartcraft.com/apps) and Microsoft Office (github.com/storytold/gridcra…) in Rust, so they are faster than the originals.
I feel "dumb" for letting Fable 5 dup me. High effort is being dumbed down to low effort behind the scenes.
Stanley Wong retweeted
We ran Kimi K3 through 3 agent harnesses (Claude Code, Hermes, Kimi Code) on 28 identical tasks.
All 3 harnesses completed the tasks at similar success rates, but the interesting story is token efficiency: the same task cost up to 30x more tokens depending on the harness. 🧵🧵
If a machine can execute your daily tasks faster, cheaper, and better than you, what actually keeps you indispensable?
It isn't faster output. It’s courage, connection, and accountable judgment.
Here is how you build a moat AI can’t touch 👇
It’s coming! Kimi K3 teaser video just dropped! Exciting to see how it performs vs. Opus 4.8 vs Fable 5 vs GPT 5.6 Sol. Rumors say near Fable 5 quality in an open source model. Wait and see!
German courts have ruled that Google is liable for false AI Overview answers. The ruling states that Google’s AI Overviews are it’s own content and not just a listing of search results. My POV: This is huge as it means lack of veracity and reliability (AI Slop) has a financial penalty associated with it! the-decoder.com/landmark-ger…
This is going to shake things up.
We've kept hearing how GLM-5.2 beats Opus 4.8, and are skeptical of benchmarks - so we tested them on a real bug from the Cline repo. While both models fixed the issue, GLM was the winner in terms of cost and code quality:
- GLM used twice as many tokens (GLM 1.1m vs Opus 660K) but cost half as much (GLM $0.41 vs Opus $0.81)
- Opus finished quicker - 1.6 min and 12 tool calls vs GLM 4.7 min and 28 tool calls
- GLM cleaned up dead code and verified the build compiled before completing. Opus didn't - it left type errors that passed tests but broke the production build.
Both runs used the same Cline harness prompting and tools, so it seems GLM is RL trained to spend more tokens verifying its work before completing. Impressive work by the @Zai_org team!
Claude is improving itself!
Our internal data shows Claude is accelerating AI development—a possible path to recursive self-improvement, or AI autonomously building a more capable successor.
It’s happening faster than we thought, and the implications deserve greater attention. anthropic.com/institute/recu…
Completely agree. Taiwan has a highly resilient semiconductor ecosystem, much like Silicon Valley has a resilient technology ecosystem. These ecosystems are difficult to replicate because they are complex, networked systems: their strength comes not from any single company or institution, but from the non-linear interactions among suppliers, talent, infrastructure, capital, know-how, and culture.
“Taiwan won’t matter in 18 months” is what happens when software people mistake civilization for a SaaS product.
It's one of the most clueless takes I've heard on this platform in months and that's saying something.
Yes, America has semiconductor fabs. Mostly old ones. Really old.
Taiwan has the fabs that matter:
The ones that make every single chip for NVIDIA and Apple and every damn Android and iPhone on Earth and even most of the 1500 or so chips that go in your truck or car.
Without TSMC these companies simply do not exist. Not kind of struggling. I mean "wiped off the freaking face of the Earth and unable to produce a single product" level gone.
As in "worth zero instantly."
Taiwan has:
- Multiple leading-edge giga-fabs
- The *overwhelming* majority of advanced AI chip production
- Dominant advanced packaging capacity
- Dense supplier clustering
- Decades of accumulated yield/process knowledge and the most skilled workforce on Earth to run it all
The US still barely has frontier-scale advanced packaging online. Much of it is literally still under construction and won’t ramp until years from now.
Momos hear “we’re only 1–2 nanometers away” and think semiconductors are just transistor geometry.
No freaking way. Sheer idiocy. The real moat is:
- Yields
- Packaging
- HBM integration
- Substrates
- Tooling
- Tacit manufacturing expertise
- Workforce density
- Supply chain coordination
TSMC is not “a fab.” It is one of the most sophisticated industrial ecosystems ever created by humanity.
And no, a tiny Neuralink surgery robot does not mean America can magically reproduce decades of semiconductor manufacturing concentration in 18 months.
Reality is not a podcast episode.
Taiwan remains strategically critical for years, likely a decade+.
This is like saying:
“We’re 18 months away from replacing the global oil system because we built a nice electric bike.”
Stanley Wong retweeted
Effective today, we are:
1) Doubling Claude Code’s 5-hour rate limits for Pro, Max, and Team plans;
2) Removing the peak hours limit reduction on Claude Code for Pro and Max plans; and
3) Substantially raising our API rate limits for Opus models.
