Why not just use ChatGPT or Google?
Generic AI is good at plausible answers. Search is good when you already know the name. Achiote is built for evidence-bounded food-memory reconstruction when the memory is fragile, partial, misspelled, or family-specific.
Generic AI or recipe search
- Often jumps from a few clues to a confident dish guess.
- Can blur what the user said, what was inferred, and what was researched.
- Usually optimizes for a complete answer or recipe.
- May miss the smallest useful sensory test.
Achiote
- Keeps User-said, Inferred, Researched, and Unknown evidence separate.
- Produces a Memory Receipt that can be corrected by the user or family.
- Prioritizes a First tiny taste test before a full recipe or drink.
- Uses feedback signals to learn whether the region, acid, texture, or difficulty was wrong.
What the Memory Receipt protects
A Memory Receipt makes the answer inspectable. The point is not to claim cultural certainty. The point is to show the trail, keep the uncertainty visible, and make the next question or bite smaller.
User-said
Words and sensory details the user actually gave.
Words and sensory details the user actually gave.
Inferred
Low-confidence context that can seed questions, not conclusions.
Low-confidence context that can seed questions, not conclusions.
Researched
Source-backed facts the host assistant found.
Source-backed facts the host assistant found.
Unknown
Gaps that should stay unresolved until someone checks them.
Gaps that should stay unresolved until someone checks them.
When Achiote wins
Use Achiote when the dish or drink name is missing, the memory comes through family language, the strongest clue is sensory, or the user needs one cheap proof bite or sip before cooking a full recipe or making a full drink. Use ordinary search when you already know the dish or drink and just need a recipe.