Issue 007 · 6 min read
Three flagship releases in one week landed at or near the price of what they replace. The lever people reach for first, pay more for better or accept less to spend less, barely moved. What is left to decide is everything that was always the actual work: whether the model changes the result on your task, what the switch costs you in migration and running bills, and where a person still checks the output before it counts.
Skip if you like
An AI assistant is a tool you talk to in plain language, like a fast, very literal colleague who has read a great deal but knows nothing about your job until you tell it. It can draft, summarise and suggest, but it does not decide, and it can be confidently wrong.
A prompt is simply what you type or say to the assistant: your request, plus any background and examples you give it. The clearer and more specific the prompt, the more useful the answer, so treat it like briefing a new starter, not typing a search.
A context window is how much the assistant can hold in mind at once: your instructions, the document you pasted, and the conversation so far. When a chat gets long it starts to forget the earliest parts, so for a big task give it the key facts again rather than assume it remembers.
Start here
Newly announced
A shared headline price removes the excuse to avoid the strong model, and it removes the shortcut of choosing by price. What is left is whether the gain shows up on your own task.
See the detail and the sources →Newly announced
The same price is not the same as free to switch. The real bill here is the migration: three breaking changes that fail quietly in production rather than at build time.
See the detail and the sources →Newly announced
Three flagships arriving at a flat price in one week is the pattern, not the exception. When price stops sorting them, the only test left is your own work.
See the detail and the sources →
The one that mattersWhat's genuinely new
Not a single model. The pattern: OpenAI, Anthropic and Google each shipped a stronger model this week without charging more for it, so a stronger option is no longer the more expensive one.
What the makers claim
Each provider frames its release as more capability at no extra headline cost. That is a price and capability claim, not evidence that the gain appears on your work, and it says nothing about the caching rates, migration effort or breaking changes that decide your real bill.
Where it helps
Choosing a default for demanding, repeated work where you can run a fixed task through the new model and score the accepted result against what you use now.
Where it doesn't
Swapping a model inside a live integration on the strength of a matching price, when forced tool use, cached prompts or reused reasoning can fail quietly once traffic moves.
What remains yours
Anthropic released Claude Fable 5.1 at the same headline input and output price as Fable 5, with prompt caching cut to about a quarter of the cost and three breaking changes for existing integrations.
Pick one demanding task you run often, score your current model on it against a fixed standard, then run the same task through this week's release and compare the accepted result, the running cost and the review time before you move any real work.
Try this
A stronger model has arrived at a price you can justify, and it is tempting to switch your default without checking whether it changes anything on the work you actually do.
Run the same fixed task through each model, summarise where the two results differ, and surface the points they disagree on.
Set the task and the standard, judge which result is genuinely better for the job, and decide whether the switch is worth its cost.
Pick one demanding task you run most weeks
Write down what a good result looks like before you run anything
Run the task through your current model and this week's release with the same inputs
Score both against your standard, and note the running cost and any migration work
Decide, and keep a person on the checkpoint before the new model's output is acted on
A matching price is a reason to test, not a reason to switch. Keep the current model in place until the new one clears your own standard on your own task.
The Human Lens · what stays yours
Price was doing quiet work: it told you the strong model was for the important jobs and the cheap one for the rest. When the prices meet, that sorting disappears, and the judgement it stood in for comes back to you. What the model costs to run, what the switch costs to make, and who checks the result before it counts were always the human parts, and this week they are the only parts left to decide.
Worth watching
One question
See you next Monday.
Bring it back to your work
Bit starts from this issue, then asks you enough about your own work to make it worth acting on. If the honest answer is that none of this touches your job yet, it will tell you that too.