The PromptFloe blog
How software gets built when you just describe it
Ideas, walkthroughs, and the thinking behind turning plain language into real, running apps.
Loop Engineering: The Playbook for Systems That Prompt Your Agents
Loop engineering is the fourth layer above prompt, context, and harness engineering — the five moves, six parts, the generator–evaluator that says “no”, the five ways loops fail, Stripe’s 1,300 PRs/week, and the four silent costs.
Read →What Is Test-Time Compute? Why Reasoning Models Think Before They Answer
Test-time compute is the computation a model spends reasoning at inference. Why o1-style reasoning models think before answering — with real benchmarks.
Read →Fable 5 vs Other AI Models: The New Mythos of Vibe Coding
Fable 5 vs other models for vibe coding: what Fable 5 is, the AI new mythos behind model names, and why the verification loop beats the model.
Read →Prompt Engineering Is Peaking. Loop Engineering Is Next.
Prompt engineering is peaking; loop engineering is next. The research, benchmarks, and patterns behind the generate-verify-reflect-refine loop that 2×–10×'s the same model.
Read →How everyone else lost at context engineering
LLMs don't fail at code for lack of intelligence - they fail on context. A research-paper look at why structural beats fuzzy, with measured proof: our compressor cuts real source 18.7× while preserving every contract.
Read →The magic loop: idea → app → ship
Building software is a conversation, not a one-shot. Here is the loop PromptFloe is built around - describe, see it run, refine, deploy.
Read →How PromptFloe works
From a single sentence to a running full-stack app - what actually happens between your prompt and your live preview, explained simply.
Read →