The way you get good at something is by doing it
What are you doing?
What are you not doing?
You are skilling up and down constantly
Show me your schedule and I will show you your skillset
Grok was dumb. This is a game where Astra killed grok by just sending a couple probes forward to kill it.
In Grok's thinking trace it was aware this was happening and considered dozens of different approaches to solving it. Do I fight back? Do I focus on upgrades to fight later? ... [then it died before it ever did anything].
pacing the frontier
pacing the moon race
pacing the nascar rally
pacing the breakfast buffet
pacing the dance floor
pacing the open bar
pacing the hallway
pacing the
the cursor essay was cool but gpt 5.6 pro also one-shotted:
* push pack WAL in s3
* fast ref CAS
* arbitrary git replica count
* geometric compaction
there's some alpha in hiring a 10y github infra vet to build your clone but even that is shrinking
crazy times
To clarify, I didn鈥檛 give 5.6 pro the cursor blog post. I asked it to improve on spokes performance and reliability and it came up with all these approaches on its own
Whether LLMs can write the code is a boring question
They have the ideas too