Multi Token Prediction is being sold as a free speed hack for local LLMs. Flip one flag in your inference engine and generation speeds up by anywhere from a quarter to a factor of three, at no cost in output quality. That pitch is accurate as far as it goes. However, there is an interesting...
Category: AI
The Org Chart Stopped Routing
A five-person organisation has always operated on a single principle, which is that there is nobody to hand it to.
You Only Know the Right Model After You’ve Used the Wrong One
Which model should run this task? The question has a correct answer. It arrives after you have stopped needing it.
Tokens are not commodities
Token pricing treats every token as interchangeable: one model's tokens against another's, dollars per unit, compare and choose.
Ten Fable prompts that turn expensive mistakes into cheap questions
The quality of the work with Fable is determined by your ability to say what you want and your ability to notice what you have not said.
You Can’t Extract Fable With a Few Prompts
A widely shared post is making the rounds with an urgent pitch: ask Fable to write down its "operating manual" while it's still free.
The jagged frontier is permanent. Build like it.
The jagged frontier is the observation that LLM capability is not a smooth surface with sometimes unexpected behavior.
The slash command that does nothing
Every few weeks a fresh list of "secret" Claude slash commands makes the rounds - in newsletters, on LinkedIn, on Twitter in Instagram reels.
Software cloning is a verification problem
Tools like Codex and Claude Code made cloning cheap in many places, but left it expensive in others where verification is hard.
The Human Is the Outer Loop
A factory, underneath the word, is a loop. The agentic software factory, the version being built right now, puts the models inside that loop.