GPT outcome-first vs Claude's surgical specificity: new LLMs punish vague prompts. Precise, outcome-focused thinking now determines quality, cost, and tokens.
Both OpenAI and Anthropic just released official prompting guides. Both say the same thing. Your old prompts don’t work anymore. But for opposite reasons. Claude Opus 4.7 stopped guessing what you meant. It does exactly what you type. Nothing more, nothing less. Vague instructions that worked on 4.6? They now produce narrow, literal, sometimes worse results. Not because the model got dumber. Because it stopped compensating for sloppy thinking. GPT-5.5 went the other direction. OpenAI’s guide literally says: “Don’t carry over instructions from older prompt stacks.” Legacy prompts over-specify the process because older models needed hand-holding. GPT-5.5 doesn’t. That extra detail now creates noise and produces mechanical output. Claude got more literal. GPT got more autonomous. Both now punish the same thing: prompts written without clear thinking behind them. One developer on Reddit captured it perfectly after analyzing hundreds of community posts. The complaints tracked almost perfectly with prompt specificity. Precise prompts got better results on 4.7. Vague prompts got worse. The model didn’t regress. The prompts did. OpenAI’s new framework is “outcome-first prompting.” Describe what good looks like. Define success criteria. Set constraints. Then get out of the way. The model picks the path. Anthropic’s framework is the inverse: be surgically specific about what you want, because the model won’t fill in your blanks anymore. Two different architectures. Two different philosophies. One identical conclusion: the person writing the prompt is now the bottleneck, not the model. Boris Cherny, the engineer who built Claude Code, posted on launch day that even he needed a few days to adjust. That post got 936 likes. Meanwhile, Anthropic increased rate limits for all subscribers because the new tokenizer uses up to 35% more tokens on the same input. The model is more expensive to run lazily. Cheaper to run precisely. The models are converging in capability. The gap between good and bad output is no longer about which model you pick. It’s about the 2 minutes of structured thinking you do before you type anything. That thinking system is the skill. The prompt is just what it produces.
Real-time analysis of public opinion and engagement
What the community is saying — both sides
Opus 4.7 stops “guessing what you meant” and will do exactly what you wrote, so remove assumed logic and be surgically specific about intent, format, and success criteria.
describe the desired result, not the step-by-step process; let the model choose the path and fill in the method.
cognitive work shifted to humans: clarity of thought matters more than clever prompt tricks.
show one good example rather than many “don’t do this” rules; match effort to task complexity; stop wasting tokens with unnecessary scaffolding like “think step-by-step” on trivial tasks.
agentic pipelines and skills that relied on previous forgiving behaviour will break or change the cost model; teams are rewriting flows and tightening specs rather than adding more instructions.
vagueness used to let models explore hidden paths and surface unexpected innovations; literalness reduces that exploratory benefit and shifts control (and potential value) back to defined prompts.
demand for prompt rephrasers, model-specific templates, and skill wrappers that convert human intent into the right format for each model is rising.
different models reward different working styles; keeping multiple models in the workspace helps match tasks to the one that best fits your preferred level of structure and ambiguity.
Several users report Opus 4.7 “takes big liberties,” “does what you ask and then 50 things you didn’t,” and even moves in the opposite direction without asking, making it unreliable for precise work.
Some argue the issue isn’t how you write prompts but how your system is built: if your stack depends on the model guessing intent, every model update breaks you; a stable constitutional/architectural layer should absorb model differences.
Others insist a truly “smart” model must understand intent, restate its understanding for confirmation, and then execute; that capability should work out of the box, not require elaborate prompt engineering.
Businesses worry that forcing a single prompting style creates a costly learning curve if they switch AI vendors, locking engineers into one way of working and reducing flexibility.
A subset accuses the vendor of intentionally crippling features (“kill Mythos’s capabilities”) and of spinning the change badly, with accusations of gaslighting and fears of monetization/mismanagement.
A few users say their prompts worked unchanged across versions and that insisting on hand-crafted prompts is a rookie move; instead, ask the model (or another model) to generate the prompt that accomplishes your goal.
Most popular replies, ranked by engagement
TLDR for the scrollers: → Claude Opus 4.7: stopped filling in your vague instructions. Does exactly what you say now. Fix: be surgically specific about intent, format, and success criteria. → GPT-5.5: stopped needing step-by-step hand-holding. Fix: describe the outcome, not
Sources: → Anthropic Claude Opus 4.7 migration guide: https://t.co/nNQqORxzAA → OpenAI GPT-5.5 prompting guide: https://t.co/iiYz0tnZ8e → OpenAI “Using GPT-5.5”: https://t.co/YGNtTyPAAY → Boris Cherny (Claude Code lead) post: 936 likes confirming adjustment period →
the whole promise of LLMs was natural language. Messy human intent in, useful understanding out … now are we making AI more human-friendly or training humans to become more machine-readable?
100% BS: "Claude Opus 4.7 stopped guessing what you meant. It does exactly what you type. Nothing more, nothing less." (I know this comes from them, not harping on you) Major reason I stopped using 4.7 is because it does what you ask and then 50 things you didn't.
You say sloppy thinking…I call bullshit. It was able to truly think about what your intent. Big difference. 4.6 was a smart model, 4.7 is a dumb tool.
re problems for businesses and enterprises ie their engineers will get locked down into one style of promoting. For whatever reason if the business decides to change the AI vendor, there will be a whole learning curve for the workforce and in AI native workplaces, ghee hardly is
Found something wrong with this article? Let us know and we'll look into it.