Build Notes

Deep Analysis, and answers that show their evidence

• Deep Analysis: switch it on in the chat box when a question needs more rigour, such as a claim, a regulation, a mechanism or a comparison of studies. Our most capable model writes the answer in one to three minutes, lists its assumptions and confidence, marks what to check before use, and explains how the answer was built.
• An evidence label on every answer: From the sources, Interpreted from the sources or AI reasoning, with a line on which sources it rests on. Deep Analysis answers label each main conclusion. The label rates where a conclusion comes from, not how certain it is.
• Allowances at a glance: Deep Analysis is included on every plan (Essential 2 a day, Professional 5, Academic 1; trials include 3). Your questions and Deep Analysis left today and this month are in the profile menu and under Settings → Plan & usage.
• A roomier chat box: the question gets the full width, with focus, tools, Academic mode and Deep Analysis in one row below it.
• Better on laptops: a slightly tighter layout on smaller screens, so your recent chats stay in view.
• The intro tour for new accounts now includes Deep Analysis, and the What's new tour shows this update.

Why: in food R&D, the difference between "the paper says" and "we think" matters. Labelling every answer makes that difference visible, and Deep Analysis gives the hard questions the time they need.

Charts in answers, and signals from your market

• Chart this table: turn any table in an answer into a chart (bars, side by side, stacked, line or scatter), switch back to the table any time, and copy either for Excel. Each chart says whether its values are calculated, taken from the sources, or an AI estimate.
• Chart your data: paste a table or open a CSV or Excel file from the chat input, see a chart straight away, then ask Alchemyst what the data shows.
• Safety signals in the Briefing: a new Trends view shows recalls and warnings by hazard (pathogens, allergens, chemical, foreign bodies) and by market, week by week, with a preview on the Briefing board and the start page.
• Market signals in the Briefing: a new Signals tab shows which ingredients, health areas, claims, themes, formats and companies trade press, trial registries and journals mention more or less often, for your product categories. They are counts of mentions, not sales data, and each one opens the items behind it.
• Relevant first: the Briefing board can now put items that mention your categories or markets at the top.
• Batch Notes analytics (Professional): compare the sensory scores of up to three batches, see each batch's pH and water activity against the EU zone where Listeria cannot grow (Reg. (EC) No 2073/2005), and follow measurements such as pH and Brix across batches.
• Library analytics and a portfolio overview of your projects.
• A one-minute tour for new accounts, a Getting started checklist on the start page, and a What's new tour for returning users.
• Usage analytics are now opt-in: an unticked box at sign-up and a switch in Settings, instead of a cookie banner.
• One consistent design for tabs and filters across the Briefing, Batch Notes and the Library.

Why: in September we promised charts inside answers. Food R&D runs on numbers, from sensory scores to pH to recall counts, and a table only gets you halfway to seeing what they mean. Market signals add the other half of the Briefing: not only what went wrong in your markets, but what your category is paying attention to.

A first dark mode, and documents in Spaces

• A dark mode toggle, the first step towards the full light and dark design that shipped in September.
• Company documents can be assigned to a Space, so the files for one project stay with that project.

Why: a quiet summer on features. Most of our time went into talking to pilot teams and planning the September redesign.

Teams

• Team workspaces: invite colleagues and work in the same Spaces.
• Branded invitation emails, so an invite looks like it comes from a product and not from a stranger.

Why: R&D is a team sport. Answers are only useful if the people who make the decision can see them too.

Open beta

• Invitations for testers from food companies, universities and startups.
• Academic accounts recognised from the university email domain, managed from an admin view.

Why: we wanted real questions from real R&D work as early as possible. Most of what shipped since came out of that feedback.

Ready for your phone

• A dedicated mobile layout, so a quick question on the factory floor or at a trade fair works as well as at the desk.
• A smoother sign-up.

Search, longer questions, papers mid-chat

• Search across all your chats.
• Room for long, detailed questions, up to 10,000 characters.
• Upload a paper in the middle of a conversation.

Why: real R&D questions come with context: a spec, a supplier sheet, a study. The tool had to make room for it.

Spaces, and the plumbing

• Spaces: save conversations into one place per project.
• A collapsible sidebar and a chat that works on mobile.
• Sign-in and account emails of our own.

Why: not glamorous, but a tool people rely on at work has to feel dependable before it can feel clever.

The first prototype

• A chat for food R&D questions, with file attachments and chat history.
• Built and tested with food scientists and product developers from day one.

Why: we had spent years watching R&D teams lose days to literature searches and regulatory lookups. We wanted to see if an assistant could give them that time back.