Changelog

After a brief rebrand, the product and site are back to the Benchmates name, with a freshly written homepage and updated visuals across the dashboard and mobile app.

The Goals and Research Jobs settings pages now let you add or remove an entry inline, right there on the page, no separate modal or navigating away and back.

When setting up a new research job, Facebook is now one of the source options alongside Reddit, Instagram, Google Trends, YouTube, and PubMed.

A new Sentiment page tracks sentiment for your brand, competitors, and topics over time, lets you backfill historical data with one click, and can generate a public, no-login link so you can share a trend with a client without giving them dashboard access.

The engine that suggests new content ideas now weighs concepts by how well they've actually performed and avoids over-relying on generic, frequently-mentioned topics, so suggestions read as more specific and less repetitive.

Bringing a new brand onto Benchmates, across the dashboard, API, and back-end setup, got streamlined into a single, more coordinated flow this week, plus a defined brand voice you can set during onboarding.

Fixed a handful of bugs causing the daily briefing email to silently fail (calendar links, delivery errors), and tidied up how it looks in your inbox.

The mobile app can now be added to your phone's home screen as a real, full-screen app, no app store needed, and it stays logged in and up to date automatically every time you open it.

Voice calls on speakerphone in a car were getting falsely interrupted by road rumble; that's fixed, so June only stops talking when you actually start talking.

Error monitoring was added across the dashboard, API, and infrastructure, so problems get flagged and caught automatically instead of quietly failing in the background.

Accepted actions can now be pushed out as drafts on Shopify and WordPress (web pages, blog posts) and shared to Instagram — the publishing platforms promised on the site are now actually wired up, not just listed.

A new Integrations page in the dashboard is where you connect publishing destinations, starting with email/newsletter delivery, instead of that being configured behind the scenes.

June and the other agents now reach a single shared connection for your Obsidian notes, Notion, calendar, and team knowledge, so she can look things up live during a conversation instead of relying only on what was prefetched beforehand.

Voice calls now support real barge-in: cutting in while June is mid-sentence stops her immediately, instead of both of you talking over each other until she finishes her thought.

June can now read multiple calendar feeds at once, understands recurring events, and can write new events directly to iCloud, not just read a single feed.

Accepted content ideas ("actions") now get a full detail page, and both the list and detail views show an AI-generated image for that specific piece — a portrait shot for social posts, landscape for everything else. No more guessing what an idea looks like before you commit to it.

A new Brand tab lets you point at your website; Benchmates analyzes it and writes a short style guide for your brand's imagery. That style guide is then applied automatically to every AI-generated image after, so you stop getting generic, off-brand-looking output.

Click any concept in the Knowledge Graph and a side panel shows the vault documents behind it, with previews, plus one-click "Ask vault" and "Actions" buttons — so you can go from an idea to acting on it without leaving the graph.

Navigation was rebuilt around Actions and Goals as the things you'd actually look at day to day, with goal tags on actions and per-goal action counts so you can see where effort is (and isn't) going.

A batch of voice-quality fixes landed: cleaner call handoffs, no more duplicated messages, and June now wakes back up automatically when new information comes in mid-conversation.

The first version of the Benchmates dashboard shipped — a home base with a Goals tab, a dedicated Actions page, and a System Health view so you can see at a glance whether everything's running. Project navigation was also reorganized to be easier to scan.

Voice conversations with June can now call in Claude for tougher questions instead of relying on the voice model alone — better answers on harder questions, without switching apps.

June now keeps her own knowledge base, so context from earlier conversations carries forward instead of resetting each time.

No new screens yet, but this week's work laid out exactly how the Knowledge Graph feature — which shows how ideas in your vault connect to each other — would be built. It shipped a few weeks after this.

June's voice conversations switched to a new voice engine, started responding faster, and picked up real call manners: mute/unmute, detecting when you've stopped talking, and speaking first when it makes sense instead of waiting around.

A live activity panel now shows what's happening behind a voice conversation as it happens, instead of a black box while you wait for a reply.

Ruby, the research agent, gained additional research capabilities this week — more groundwork for the content research pipeline.

June's mobile voice/chat client got a full rework of how messages move in real time, swapping an older relay for a more reliable event system. Nothing new to click yet, but it's what made the smoother voice conversations in the following weeks possible.