Brian Gingras @briangingras - Twitter Profile

about 1 month ago

After yesterday’s Grok Bot Galaxy Day 1, these are the four use cases I think are genuinely useful. Together, they let a Bot team spot opportunities, help you decide what to do, follow through on the work, verify the results, and share corrections. The first already has a template you can import. 1. Competitor Research Bot Import the Stalk Bot template Shub shared and tell it about your product and which competitors to watch. It uses a dedicated research email to subscribe to competitor newsletters and sign up for products, records the onboarding process, and monitors websites, pricing, changelogs, hiring, and X. It then compares what it finds with your product’s positioning and suggests how to respond. Have it establish a baseline before setting a monitoring schedule. Without a previous record, a finding can only be marked “first seen.” It needs a before-and-after comparison to call something a change. The template also sets boundaries: no public posts and no outreach to competitor employees or customers. 2. Coordination Bot Give your coordination Bot your priorities for the week. Have it use connected email, work messages, and meeting notes to track what you’re actually working on, each project’s status, and what needs to happen next. Then compare that with your original plan. This is the “attention list” approach in the official PM guide. For example, if launching is your top priority but you’ve spent two days answering messages and handling interruptions, have it flag the launch work that hasn’t moved. New competitor findings and customer feedback can also be checked against this list, with decisions that need your input pulled out separately. If nothing important has changed, it stays quiet. 3. Engineering Bot If you run coding agents, follow the approach in the official engineering guide: have an engineering Bot handle delegation and follow-up checks. It checks PRs, test results, review comments, and merge conflicts, sends problems back to the coding agent, and updates the task’s status. Work goes to review once it meets the agreed criteria. Define the completion criteria and required evidence when assigning the task. UI changes need screenshots of the actual page; functional changes need test and runtime results. When a check fails, have it send the work back with the specific problem to fix. Agree in advance on what the Bot can resolve independently and when it needs you. Set merge and release permissions separately. 4. Team Corrections Have the coordination Bot document your corrections in a shared playbook, specify which tasks they apply to, and notify the relevant Bots. The official engineering guide uses this approach too: investigate the mistake, update the playbook, and share the correction with the other engineering Bots. For example, if a type of task must include test results and screenshots, pass that requirement to every role handling that work. You shouldn’t have to explain it to each assistant separately. Keep temporary, task-specific instructions out of the team’s permanent rules. The four can work together like this👇 The research Bot finds a competitor change and provides before-and-after records and recommendations → The coordination Bot checks it against the current plan and lists the decisions that need your input → You approve the changes to make → The engineering Bot delegates, checks the results, and sends failed work back → Corrections worth keeping go into the shared playbook and are passed to the relevant Bots If you already have a team, you don’t need four new Bots. Give competitor research to the research role, the attention list and team corrections to the coordinator, and execution follow-up to the engineering role. Get the workflow working once first. Save repeatable methods as Skills and scheduled work as Routines. Create a separate Bot when the work needs its own ongoing task queue, records, and regular handoffs. Keep confirmation requirements for external messages, publishing, and other actions that need approval. Stalk Bot template: https://t.co/gJKu6WcpMx

4

195

20

344

16K

BrianGingras retweeted

Whole Mars Catalog

@wholemars

about 2 months ago

Tesla Cybercab launches today

64

3K

175

31

76K

BrianGingras retweeted

Sawyer Merritt

@SawyerMerritt

about 2 months ago

Next week will mark the first time that a purpose built Robotaxi (Cybercab) with no steering wheel or pedals that can be mass produced will offer rides to members of the general public in North America, and 99% of the population has zero clue this is about to happen.