Veelgestelde vragen

• Mention any fact in a conversation — «I’m vegetarian», «Mom’s birthday is March 5», «I use React at work»
• Thremark automatically extracts it and creates a mark — you’ll see it in the Marks tab with a category and timestamp
• In future chats, your marks are loaded silently — ask for a meal plan and it already knows your diet, no reminder needed
• Edit, delete, or add marks manually anytime — you’re always in control of what Thremark remembers

• Memory — ChatGPT: saved memory is largely opaque; you see hints and summaries, not every fact the model may use. Thremark: what the AI learns becomes visible marks—searchable, editable, deletable—so you always know what context is in play.
• Platform — ChatGPT: a strong cross-platform chat app. Thremark: native SwiftUI on iPhone with deep Apple integration—Siri, Dynamic Island, widgets, Share Extension, Spotlight, and dozens of system hooks—not a generic wrapper around a single web chat.
• Models — ChatGPT: you use OpenAI’s model lineup inside OpenAI’s product. Thremark: Smart Routing picks Gemini, GPT, or Claude per turn from your vibe (Fast / Smart / Creative) and plan—you don’t micromanage provider APIs.
• Workflow — ChatGPT: conversation-first; other capabilities are layered on. Thremark: chat, structured Files, web search with citations, image generation, and Imprint voice share the same marks-backed context in one product.
• Privacy & training — Both send prompts to LLM providers to generate replies. Thremark does not use your content to train its own models and does not sell your data (see the dedicated FAQ on training below).

Other assistants may remember things too, often as an opaque summary. Thremark is built around marks—every saved fact is visible, searchable, and editable. Typical vendors vary; this table shows our product shape. A dash (—) means rarely or not at the same depth as native Thremark.

• Ask Thremark to create something structured — «make me a weekly workout plan» or «track my monthly expenses»
• A named file appears in your Files tab — formatted, organized, and ready to use
• Open it anytime to view or edit — or ask Thremark in chat to update it («add yoga on Fridays»)
• Because Thremark knows your marks, files are personalized from the start — a meal plan respects your diet, a travel checklist knows your destination

• Describe what you want — «landing page for my studio with booking and pricing»
• Thremark writes code and shows a live preview — you watch it take shape
• Refine by talking — «tighter typography», «add a dark mode toggle» — no manual file juggling
• Deploy when it’s ready — you own the code, no platform lock-in

• Describe what you want — professional («app icon, flat, two colors») or personal («portrait, natural light, warm tones»)
• Thremark generates the image right in the chat — Free and Plus use Gemini Flash; Pro uses Imagen 4 Ultra at 2K resolution
• Iterate — «softer background», «more contrast», «try black and white» — until it feels right
• Save to your device, or take the concept to Vibe Code if you’re building something bigger

• Model — Free and Plus: Gemini 2.5 Flash Image (fast, good quality). Pro: Imagen 4 Ultra — Google’s highest-quality image model with 2K resolution output.
• Limits — Free: 90/mo. Plus: 150/mo. Pro: 300/mo (Imagen 4 Ultra).

• Tap the microphone icon in any chat to start a voice session
• Talk naturally — Thremark listens in real time and responds with voice, using your full memory context
• Everything you can do in chat works by voice — save facts, create files, search the web, ask questions
• End the session anytime — new marks and files created during the conversation are saved automatically

• Providers — Generic assistants: often push-to-talk or text-only. Thremark Imprint (iOS): duplex streaming to Gemini Live or OpenAI Realtime with your marks in the same session context as chat.
• Web — Browser microphone sends speech-to-text into the chat input; full duplex Imprint with streamed audio is iOS-only.
• Storage — Voice audio is processed by the provider for the reply and is not stored on Thremark servers as a recording (see Privacy).

• Ask anything that needs current data — «best noise-cancelling headphones under $300» or «what happened in tech today»
• Thremark searches live sources, reads them, and synthesizes an answer — not from training data, from the actual web
• Every fact comes with a source link you can tap to verify — cited, not guessed
• Because Thremark knows your marks, search results are shaped by your preferences — it knows your budget, your tech stack, your taste

• Grounding — LLM-only answers: risk of stale or invented facts. Thremark: retrieval-first, then synthesis with mandatory source links so you can verify.
• Scope — «Web Search» here is the in-chat product feature (internet lookup), not the thremark.com web app (see Platforms).

• Choose a vibe — Fast, Smart, or Creative — instead of picking a provider or model name
• Each message is routed server-side to Gemini, GPT, or Claude based on the task, your plan, and availability
• Your marks and chat context travel with the request — the model choice changes; your memory does not

• Control — Many apps: pick one vendor’s model list or a single default model. Thremark: one vibe control; routing picks the engine per turn so you don’t micromanage APIs.
• Plans — Free emphasizes Fast; Plus unlocks Smart; Pro unlocks Creative and the strongest routing tier within your limits (see Pricing).

• Providers — Third-party LLMs: each vendor has its own terms for API traffic; Thremark does not add your chats to a proprietary training dataset on our side.
• Product — Marks and messages are for your product experience only — not sold, not used to train a separate «Thremark model.»

• Stack — Web: Next.js client to the same Worker API as iOS; full cloud sync for marks, files, and chats. Native-only features (Imprint duplex, Siri, widgets, Dynamic Island, etc.) are not available in the browser.
• Billing — Subscriptions for Plus/Pro currently go through Apple on iPhone; the web app respects the entitlements tied to your account. Web billing is planned.