Issue 008 · 6 min read
Each new release puts more on autopilot, but very little of your own oversight can go away. If a Bot or agent keeps working after you clock off, your unfinished drafts, open downloads or logged-in accounts persist until you reclaim or clear them. Fancy benchmarks can't tell you if your buried spreadsheet formulas or client attachments are safe. The more a new tool promises, the more important the specifics of what it leaves behind. Getting it done isn't hard, knowing exactly what's done is the work.
Rather listen?
Skip if you like
An AI assistant is a tool you talk to in plain language, like a fast, very literal colleague who has read a great deal but knows nothing about your job until you tell it. It can draft, summarise and suggest, but it does not decide, and it can be confidently wrong.
A prompt is simply what you type or say to the assistant: your request, plus any background and examples you give it. The clearer and more specific the prompt, the more useful the answer, so treat it like briefing a new starter, not typing a search.
A context window is how much the assistant can hold in mind at once: your instructions, the document you pasted, and the conversation so far. When a chat gets long it starts to forget the earliest parts, so for a big task give it the key facts again rather than assume it remembers.
Start here
Newly announced
Astra gives you the option to send over bigger jobs, but the bar for trusting results hasn't moved. You have to check specifics every time, especially where the model might act without checking back with you.
See the detail and the sources →Newly announced
Grok 4.6 can run some tasks longer and a bit differently, but swapping in a new model isn't automatic. Test it on your own spreadsheet, not just on today's marketing.
See the detail and the sources →Newly announced
Grok Bot lets your work run when your computer is off, but your Bots all live in one account. If you forget to sign out or tidy up, your handover files and browser sessions are still there for every Bot you own.
See the detail and the sources →
The one that mattersWhat's genuinely new
Astra's launch expands what agent-like tasks you can hand over, Grok 4.6 tweaks its performance on long jobs, and Grok Bot lets tasks persist and overlap in one pooled environment.
What the makers claim
OpenAI's naming Astra universal and critical on cyber capability, xAI declaring Grok 4.6 their fastest and most intelligent, and both firms' benchmark talk.
Where it helps
When you set workflows to run overnight and come back in to pick them up, or when you delegate a document redraft and need to know what files are being touched.
Where it doesn't
One-off tasks you supervise live, where the work happens in front of you and nothing sensitive stays running in the background.
What remains yours
OpenAI started rolling out GPT-6 Astra on 3 September 2026, initially to a limited set of organisations and, over the following days, to ChatGPT Plus, Pro, Business and Enterprise users, the OpenAI API and AWS. It is aimed at research, coding, computer use, document, spreadsheet and presentation work and long-running agentic tasks. The API model id is gpt-6-astra, with a 1,050,000-token context window, up to 128,000 output tokens, a knowledge cutoff of 30 April 2026, text and image input with text output, and adjustable reasoning effort. Its headline API pricing, $10 per million input tokens and $50 per million output tokens, matches Claude Fable 5.1, while the two differ on cache economics, where Fable 5.1 reads cache at $0.25 per million tokens against Astra's $1. OpenAI says Astra is the first model it has rated at the Critical cybersecurity capability level under its Preparedness Framework, its stated reason for the phased rollout.
Check for any still-open credentials or personal files the Bot or agent left behind.
Try this
Find and close the loose ends before your day ends.
AI writes a brief checklist of open browser tabs, downloads and signed-in accounts it has touched in your last session.
You confirm which accounts to sign out of and which files to tidy or delete.
Have your agent write a summary of everything it accessed or changed: files, tabs, credentials.
Review the summary and flag anything you don't want left open or running.
Manually close accounts, delete downloads or move files as needed, don't assume the agent did it for you.
Check for any still-open credentials or personal files the Bot or agent left behind.
The Human Lens · what stays yours
However much you automate, there are handovers, logins and stray documents that don't clean themselves up. Knowing what's still sitting there with your name on it is the quiet task that belongs to you alone, and it's the easiest to forget until it bites.
One question
See you next Monday.
Bring it back to your work
Bit starts from this issue, then asks you enough about your own work to make it worth acting on. If the honest answer is that none of this touches your job yet, it will tell you that too.