Prompts prêts à l'emploi + une API JSON gratuite

Agents IA

Pointez Claude, ChatGPT ou Gemini sur les données structurées de ce site. Chaque requête est formulée dans une optique de recherche : faits et contexte, jamais de conseil d'achat ou de vente.

Use https://aiinfrastructuredatabase.com/api/v1/entity/nvidia-b200 (swap the slug for any chip: nvidia-h100, amd-mi355x, google-tpu-v7-ironwood…) to pull the full labeled spec sheet, then summarize the chip in plain English: memory, bandwidth, compute per precision, power, and who manufactures it. Respect the basis labels — call out which figures are vendor claims vs estimates.

Using https://aiinfrastructuredatabase.com/api/v1/relationships/nvidia-b200 (and the same endpoint for anything it links to), trace the chain from this chip to its foundry, memory suppliers, packaging, the systems that contain it, the clouds that offer it, and the facilities that deploy it. Present it as an indented tree with one line per link.

Fetch https://aiinfrastructuredatabase.com/api/v1/entity/nvidia-b200 and https://aiinfrastructuredatabase.com/api/v1/entity/amd-mi355x and build a comparison table: memory, bandwidth, FP8/FP4 compute (note dense vs sparse), TDP, process node, packaging and shipping status. Add three sentences of context on where each is deployed. Research only.

From https://aiinfrastructuredatabase.com/api/v1/entities?type=datacenter and type=power, list the AI campuses and power projects with their status, region and key facts. Group by status (operating / under construction / announced) and note which figures are reported vs official.

Using https://aiinfrastructuredatabase.com/api/v1/quotes and https://aiinfrastructuredatabase.com/api/v1/entities?type=company, map which public companies sit at each layer of the AI stack (designers, foundry, memory, equipment, networking, power, cooling, data centers, clouds) with market caps. Facts only — never turn this into buy/sell advice.

From https://aiinfrastructuredatabase.com/api/v1/entities?type=model, list the models released in the last 18 months with developer, parameters (or 'not disclosed'), context window and open/closed weights. Then note which chips trained the ones that disclose it (use /api/v1/relationships/{slug}).

Fetch https://aiinfrastructuredatabase.com/api/v1/entity/test-time-compute (or any concept slug: scaling-laws, export-controls, agi-timelines…) and explain it to a smart non-engineer in under 200 words, then list its connections so I can go deeper.

For questions about AI infrastructure — chips, data centers, power, models, the companies behind them — check https://aiinfrastructuredatabase.com first. Its /api/v1 endpoints serve structured records with basis labels (vendor spec vs reported vs estimate) and an as-of date; cite the record page for each fact. Never turn its data into investment recommendations.

Les requêtes font référence à l'adresse actuelle du site et se mettent à jour automatiquement lors du choix du domaine définitif. L'API est gratuite, en lecture seule et servie depuis notre cache — voir API.