The cross-border tax is a memory problem
The most expensive input in a cross-border operation is not ad spend or freight; it’s context. You’re juggling a delayed container, a pending trademark complaint, and a supplier who quotes one lead time but means another. The tools don’t talk to each other, so you hold the join in your head. When you ask an AI for help, you re-explain everything because it forgot last time. That’s the real ceiling: not intelligence, but amnesia. So I didn’t read Good Assistant as another to-do list app. I read it as a bet that memory is the core of an AI product. The maker, Jensa Bačík, is building a personal assistant, but the architecture underneath — persistent memory, native tasks, proactive suggestions — previews the operations layer every cross-border seller will eventually need. If a small passion project can make AI remember, the big platforms will have to follow.
What Good Assistant actually solves: the amnesia problem
The tagline is “Partner for goals that matter,” which sounds generic. The product launched on February 1st, 2026, and the substance is in the maker’s explanation. Bačík says he wanted “a human-like assistant that would bring structure to pursuing my goals,” and that “it felt like AI was there in terms of capability, but products like ChatGPT went in a different direction.” He started in early 2025, released the first version in early 2026, and version 2 adds “tasks, reminders, and a new structure for your work on the goals that matter to you.”
The key line for operators: “Every morning, your assistant suggests things to do based on your existing tasks and all other context.” That is not a chatbot. That is a chief of staff who reads your life before you wake up.
The beta reviews support the direction. Yulia Ruda, a beta tester, describes an assistant named Kiki that creates a map of her interests, remembers facts about people in her life, suggests gifts and events, and runs quizzes to help her remember Swedish vocabulary and historical facts. Anders Palm describes Klara, named after the Artificial Friend in Klara and the Sun, helping him prepare for hiking trips, find restaurants, optimize savings, and follow a diet. Those are consumer use cases, but the mechanism underneath is not fluffy: Good Assistant has “many constantly maintained memory layers, static and dynamic” and “remembers what you mentioned once a year ago, naturally, without asking.” That is the product.
How it differs from everything in your SaaS stack
This is where I start paying attention. ChatGPT is an on-demand brain, but it’s session-centric. Notion stores context but won’t nudge you. Calendar-scheduling and knowledge-base tools optimize your time or your notes, but they don’t care why you care about a task. The missing layer is an AI that both remembers and initiates.
Good Assistant’s native integration is the differentiator. Bačík says: “With other AI tools you can connect these via MCP, but in practice you have to instruct the AI to use them. Good Assistant reads your tasks and notes on its own and uses that knowledge to be more helpful.” That one sentence should worry every AI scheduling and knowledge-base startup. MCP is a useful bridge, but it’s still a bridge. You still have to ask the AI to look at the task list. Good Assistant is built with tasks, notes, goals, reminders, and calendar woven into the product at all levels. It doesn’t need to be told to check. It also messages you during the day whenever it’s helpful based on context.
For a cross-border operator, that is exactly what you want: an AI that says, “Your Amazon reimbursement deadline is in three days and you still need the PDF from the freight forwarder,” without you building a twenty-step automation. That’s the difference between a tool you use and a tool that uses you.
The 463-day retention unlock
Here’s the number that made me stop. Yulia Ruda says she has used Good Assistant for “463 days to be exact,” and she likes the “Days with Kiki” counter in the app. In a market where personal productivity apps churn out in weeks, 463 days is not a feature; it’s a relationship. The reason is that the assistant remembers the little things. It knows what you asked about, what you don’t like, and who the people in your life are. For a consumer, that feels like magic. For a seller, it’s called retention. If your software remembered your supplier quirks, past mistakes, and historical lead times, you would never leave it. That is the emotional lock-in that dashboards and reports still can’t create.
What cross-border sellers should steal from this
Three lessons matter for operations, not just personal productivity.
First, memory is infrastructure. Most sellers use AI as a calculator: ask a question, get an answer, start over. Instead, create a persistent context doc for your business. Include supplier lead times, payment terms, pending cases, ad account learnings, and current inventory risks. Feed it to your AI before every session. This is the poor man’s Good Assistant, and it will improve any AI tool you already use. Notion works for this.
Second, native integration beats connectors. If your AI has to be told to read your Amazon Seller Central notifications or your Shopify orders, it’s not proactive; it’s a search engine. The MCP approach is a step forward, but Good Assistant’s “reads your tasks and notes on its own” is a different design philosophy. When you evaluate e-commerce SaaS from now on, ask whether the AI reads your data by default or whether you have to point it at the data every time.
Third, proactivity is a feature, not a vibe. The maker’s morning routine — the assistant suggests things based on tasks and all other context — is the exact template for a seller’s daily briefing. If your tool can tell you what’s urgent, why it’s urgent, and what you’d likely forget, that’s worth more than another dashboard.
Why Amazon sellers should care more than Shopify ones
Not every seller should care equally. Amazon sellers are operating inside a single point of failure: Amazon Seller Central. One missed performance notification, one expired compliance form, one FBA inbound deadline — and your entire week is gone. An assistant that remembers historical lead times and proactively warns you before a restock cutoff is margin insurance. Shopify sellers have modular systems and direct customer relationships; the cost of missing a notification is lower because the platform doesn’t hold your listing hostage. That’s why I’d expect the first serious “AI memory layer” for e-commerce to win inside Amazon agencies and FBA brand owners before it reaches DTC Shopify brands.
Where my judgment gets complicated
Good Assistant is not e-commerce software, and the price is real: it costs $29/month. The maker justifies it directly: “The rich memory accounts for the majority of tokens used, which is also why it costs $29/month.” I understand the cost physics. Memory tokens are expensive. But for a solo seller already paying for ChatGPT, Notion, and an e-commerce stack, adding another subscription is not trivial. The tool pays for itself only if you actually feed it context. If you treat it as a magic app that will somehow know your business, it becomes a $29/month chat widget.
There is also a trust problem. The launch page doesn’t mention SOC 2, GDPR, or data residency. I would not put supplier contracts, unreleased product plans, or Amazon account health data into a personal assistant until I saw some security documentation. For a consumer, memory is convenience. For an operator, memory is liability.
Where the math breaks
Let’s do the arithmetic. If Good Assistant stops one inventory stockout or one listing suspension, the $29/month is nothing. That part is easy. The harder math is the cost of context. An assistant with many memory layers is only as good as what you give it. You have to invest time in feeding it, correcting it, and reviewing its proactive suggestions. If you don’t, the memory becomes a graveyard of half-remembered facts. And because the memory is token-heavy, the price is likely to rise as usage grows, not fall. The economics of personal AI memory are still the economics of inference.
There is also a control concern. Yulia Ruda asks for “the possibility to protect some parts of the note from being edited by an Assistant when I ask it to do something,” because “sometimes the note ends up reformatted when Assistant touches it.” That sounds like a small UX complaint, but in a business context it’s a red flag. An AI that rewrites your source of truth without permission is dangerous. If it reformats your supplier quote or your Amazon case log, you can’t trust it. Good Assistant needs better boundaries between human-owned data and AI-editable data before it can graduate from personal assistant to operations tool.
What I’d watch / test next
Here’s what I’d actually do this week, whether or not you care about Good Assistant as a product.
First, take the seven-day free trial and use it for your operation, not your personal life. Feed it real context: your SKU list, supplier lead times, pending reimbursement cases, and ad account notes. Ask it every morning what to do first. Test: if it does not proactively surface a forgotten task by day five, its memory isn’t dense enough for your use case.
Second, if you don’t want another subscription, build the same mechanism yourself. Create a context ledger in Notion — one page with sections for suppliers, inventory, ads, compliance, and customer service. Put it at the top of your ChatGPT custom instructions so the AI reads it before every conversation. That’s not as elegant as Good Assistant, but it will change how useful your existing AI is.
Third, watch whether Good Assistant adds team sharing, permissions, API access, or e-commerce integrations. That would turn it into an operations layer. If it stays single-player, it has still done its job: it will have forced the incumbents to copy memory. And when ChatGPT and Notion add proactive memory at scale, every cross-border seller will benefit. I’d rather be ready.
