Slackbot Can Build the Dashboard Now
Slackbot builds live dashboards and runs code, Gemini drafts presentations and videos, and Grok 4.5 gets a measured coding test.
What matters today
Build a Slackbot Surface, draft presentations and videos with Gemini, and test Grok 4.5 on cost per accepted coding task in PromptHacker Issue 137.
Key points
- Confirm the workspace has Enterprise+ and the new Slackbot features.
- Name the audience, decision, controlling source, and approved reference deck.
- Choose a completed bug fix with a known patch and passing tests.
- Run the same completed assignment once with the instruction and once without it.
- Choose a report already discussed in one channel.
Welcome byte
Slack is making a serious attempt to turn the channel into the place where work gets assembled, checked, and maintained. Its new Slackbot can build a live Surface, run Python, draft slides, and generate images. I am interested in the Surface more than the spectacle. A pinned dashboard that stays current could remove one of those Friday rituals everyone tolerates and nobody values.
This issue tests that idea in two layers: a broad guide to Slackbot's new Advanced Features and a ten-minute dashboard setup. Gemini gets the same treatment for presentations and video, while Grok 4.5 gets a cost-per-accepted-task benchmark instead of a launch-day victory lap. The through line is simple: let AI do the assembly, then make the source, review, and stop point impossible to miss.
Quick Hits
The reported GPT-5.6 price cut does not match OpenAI's official rates
OpenAI's official pages still list Luna at $1 input and $6 output and Terra at $2.50 input and $15 output per million tokens. Read more
ChatGPT Voice reached Work and Codex for Business desktop users
Business users on macOS and Windows can start, interrupt, and coordinate a Work or Codex task by voice within existing permissions. Read more
Gemini in Docs added 11 languages
Google expanded drafting, rewriting, summarizing, and comment help across 11 additional languages with a gradual rollout. Read more
Grok 4.5 costs $2 input and $6 output per million tokens
The xAI API rate is attractive for coding and agentic work, but accepted-task cost should include retries and review time. Read more
Meta Muse Image is live while Muse Video remains a preview
Muse Image is available in selected Meta surfaces; Meta describes Muse Video as a preview coming later. Read more
Top AI Updates
1. Slackbot can turn a channel request into a working artifact
Slack released Surfaces, Code Execution, Slide Creation, and Image Generation on July 29. Surfaces can hold a pinned report, tracker, comparison, or dashboard. Code Execution runs Python in a secure sandbox. Slide Creation and Image Generation handle the presentation layer. The Advanced Features launch on Enterprise+ with no usage limits initially.
Why it matters: I can see a team replacing one manually rebuilt status report without moving the discussion into another tool. The plan limitation and the review burden are real: a Surface can repeat a stale source just as efficiently as a current one.
Action items:
- Confirm the workspace has Enterprise+ and the new Slackbot features.
- Build one static Surface from a report whose correct answer is known.
- Check the source, calculation, channel audience, and review state before pinning it.
2. Gemini can draft the full deck and its video companion
Google's July Workspace Feature Drop adds complete presentation generation in Slides and a Gemini-assisted video workflow in Vids. Slides can use Docs, Sheets, PDFs, or existing decks as references, and an approved deck can guide the visual style. Vids can add a digital presenter in eight languages.
Why it matters: I would use this to remove the blank-deck work, then spend the saved time on the argument and numbers. Google says the Slides workflow is English-only on eligible Business and Enterprise Standard and Plus editions, plus Google AI Pro and Ultra, so account checks belong before the deadline.
Action items:
- Name the audience, decision, controlling source, and approved reference deck.
- Review the outline before accepting the slide layouts.
- Create the video from the reviewed deck so both formats keep the same claims.
3. Grok 4.5 needs a cost-per-accepted-task test
xAI launched Grok 4.5 on July 16 through its API, Grok Build, and Cursor on all plans. The API rate is $2 per million input tokens and $6 per million output tokens. That is the easy number. The harder one includes retries, review time, and the percentage of patches a team can accept without major repair.
Why it matters: I would give Grok a resolved, low-risk coding task with deterministic tests, then compare it with the current default under the same brief and environment. One cheap successful run is not a routing policy.
Action items:
- Choose a completed bug fix with a known patch and passing tests.
- Log reasoning effort, tokens, elapsed time, review minutes, and scope violations.
- Give the model a narrow default lane only after three comparable results.
Pro Tip
Add a plan-and-check pass before the first draft
I am using this on proposals and analyses where the wrong frame is more expensive than an extra minute of model time: Before you answer, work through four steps silently: understand what I actually need, sketch two possible approaches and pick the stronger one, do the work carefully, then check your own answer against my original request before you reply. Give me the final answer plus a short list of what you checked.
The important addition is an observable acceptance test. I tell the model which sources control, what the draft must contain, what it must avoid, and how long it can be. Its check list becomes a review aid, not proof that the checks passed.
Action items:
- Run the same completed assignment once with the instruction and once without it.
- Verify the claimed checks against the real sources.
- Save the prompt only where it reduces errors or review time.
Productivity Gem
Build one self-updating dashboard inside Slack
My first Surface would have five fields: project, owner, milestone, status, and blocker. I would tell Slackbot which tracker controls the facts, use channel messages only to explain blockers, and label conflicts Needs review. After two clean manual runs, I would save the recipe as a Skill and pair it with the Surface for automatic updates.
The research estimate says this can save about an hour a week. I would measure it. Count the old collection, spreadsheet, review, and reposting time, then subtract the new review, exception, and maintenance time. The difference is the actual gain for that team.
Action items:
- Choose a report already discussed in one channel.
- Verify a green, yellow, and red row before pinning the Surface.
- Publish the source and last-updated time beside the dashboard.
Health Tip
Bring three Apple Health trends to a checkup, not the raw export
Apple can export the full Health record as XML, but I would not upload that archive to a general-purpose chatbot for this task. I would review the last 90 days in the Health app, manually transcribe three relevant trends, remove identifiers, and ask Gemini to organize neutral questions for a clinician. The prompt should prohibit diagnosis, treatment advice, and medication changes.
The one-page brief should show the observation, relevant context, missing information, and question. Tracking gaps, device changes, travel, and illness belong beside the numbers. Gemini can help prepare the conversation. A qualified healthcare professional interprets the data and decides what happens next.
This information is for educational purposes only and is not medical advice. Please consult a qualified healthcare professional before making changes to your health routine.
Kids Tip
Train a NASA-style voice-command classifier
A child can use Teachable Machine's Audio Project to record "launch," "pause," and "home," train a classifier, and test it from different distances and with ordinary room noise. The misses are the lesson. They show that the model learned patterns from labeled examples, not the meaning of a command.
An adult should review microphone permission, keep personal words and background conversation out of the samples, and avoid publishing recordings. The build works without another AI account. A harmless family game is enough: the prediction tells someone to launch, freeze, or point home.
Wrap Up
I am going to test Slackbot on the least glamorous report I can find. If it can keep that one current, show its sources, and survive a month of review, it earns the next workflow. Reply with the dashboard your team still rebuilds by hand.
About the author
Pierre has spent 25+ years building growth systems across fintech, real estate, lending, campaigns, and AI workflows, with $1.5B+ in client value delivered.
Email usThree deep dives. Four useful moves. One email worth opening.
PromptHacker turns the AI firehose into practical next steps for work, health, family, and everything time keeps trying to steal.