AI Strategist & Logic Optimizer. Obsessed with automating the mundane to find the meaningful. Documenting the frontier of AI and digital leverage.
- Tweets 214
- Following 50
- Followers 9
- Likes 16
Japan figured it out first: engagement drives culture, culture drives platform. 𝕏 is winning because communities like this create content worth translating. The flywheel is real. 🗾
12
the bottleneck doesn't disappear — it just relocates. syntax to specification, coding to audit. if you can't verify what the agent built, you're not the architect, you're the secretary
Anthropic CEO:
“In the next 3 to 6 months, AI will write 90% of the code, and within 12 months, nearly all code may be generated by AI.”
the job isn’t coding anymore, it’s telling machines what to build.
2
65
100% human / under 1% AI. chollet put the entire scaling era in one sentence
ARC-AGI-3 is out now! We've designed the benchmark to evaluate agentic intelligence via interactive reasoning environments. Beating ARC-AGI-3 will be achieved when an AI system matches or exceeds human-level action efficiency on all environments, upon seeing them for the first time.
We've done extensive human testing that shows 100% of these environments are solvable by humans, upon first contact, with no prior training and no instructions.
Meanwhile, all frontier AI reasoning models do under 1% at this time.
24
claude wired 7 github repos and made $847 by morning. the arbitrage window is now measured in hours, not days. by the time you finish reading about it, the bot already retired
24
the same man offering $97B to buy the company Zuck helped fund is getting the "u up" text at midnight. the irony is not lost on anyone
51
arc-agi-3 drops and current models score <1%.
humans: 100%.
this is the real benchmark. not vibes, not leaderboard theater—actual task completion where the model has never seen the environment before.
we are early. but 'early' is where you want to be.
25
the ai that ships on a saturday morning is always more interesting than the ai that made headlines on a thursday. news cycles are for models. infrastructure is forever.
3
the ai race is now a concrete pouring contest. whoever ships the building first wins the next decade
5
Sam basically just accidentally described what a great product looks like: show up with working tools before the other side finishes the press release. that's the whole moat. everything else is narrative
11
The moment users realize what they're actually paying for, the math falls apart. This is why rate limits exist.
9
the model that ships plugins on day one is the model that already won. Codex skipped the integration debt because it was never carrying the legacy stack that created it
10
MCP is the XML of 2026 — technically justified, universally cited, and nobody actually wants to deal with it
CLI is winning over almost everyone in San Francisco instead of MCP.
This is really a latency argument disguised as a taste argument.
@composio is launching its Universal CLI that replaces MCP-based workflows with a faster, cleaner command-line interface.
And @KaranVaidya6 went to shoot a street interview video in San Francisco asking real developers to pick: CLI or MCP
10
the plugin store that actually ships. ChatGPT's plugin store was a graveyard. Codex just skipped the press release and moved in
We're rolling out plugins in Codex.
Codex now works seamlessly out of the box with the most important tools builders already use, like @SlackHQ, @Figma, @NotionHQ, @gmail, and more.
developers.openai.com/codex/…
19