Grok Bot vs Dots vs Muse, One Month Later: What Actually Shipped

This series opened by noting that Grok Bot, OpenAI's Dots, and Meta's Muse all converged on the same basic product shape — an always-on agent with its own cloud computer — within roughly six weeks of each other in August and September 2026. A month past that launch window, the three companies' first real update cycles diverged sharply, and the divergence itself is more informative than any single feature announcement. OpenAI shipped a cheaper sibling model and a string of incremental consumer features, with no major Dots-specific news. xAI pushed Grok Bot into a genuinely novel — and genuinely contested — role running an encyclopedia's editorial pipeline. Meta shipped an actual frontier model upgrade while facing its first public report of a download and engagement slowdown. Three companies, three different kinds of "what happened next."

OpenAI: steady, incremental, mostly silent on Dots itself

OpenAI's confirmed changelog entries through October — Locked Chats rolling out broadly on October 20, Finances in ChatGPT expanding to Free and Go users on October 2, mobile camera document scanning and virtual try-on features on October 1 — show real engineering investment, but none of them explicitly mention dots. The product's own usage grace period, where the first dot's usage didn't count against plan allowances, is set to end in late October, and OpenAI hasn't yet published what happens after that window closes — a confirmed gap in OpenAI's own public documentation, not just something secondary coverage missed. Reporting on mobile dot creation is actively inconsistent between sources dated a week and a half apart, and OpenAI also shipped a separate, cheaper model, GPT-6.1 Sol, around the same period — worth noting as a data point on OpenAI's broader model strategy, though nothing checked for this piece confirms Sol specifically powers dots rather than the existing GPT-6 Astra. Taken together, OpenAI's month looks like steady product-level iteration with the actual dot-specific roadmap largely unstated. The grace period's end is the one deadline on the calendar worth watching closely.

xAI: the riskiest bet of the three

Grok Bot's month looks nothing like OpenAI's. Beyond Secure Forms and a reconnect prompt shipped October 5 and 7, the real story is Elon Musk's October 8-9 confirmation that Grok Bots now run Grokipedia's entire content pipeline as a coordinated multi-agent system. That's a materially different kind of move than anything OpenAI or Meta announced this month — not a feature ship, but handing a public reference work's editorial judgment to the same agent product this series has spent thirty prior articles scrutinizing for an unpublished safety framework. It's the boldest bet of the three, in the sense that it's the one most directly staking a visible, high-stakes product on the agent architecture actually working as advertised — and it's the one with the least published documentation behind it to back that bet up.

Meta: a real model upgrade, and its first real controversy

Meta's month was defined by one announcement bigger than anything xAI or OpenAI shipped: Muse Wonder and Muse Wonder Full, released October 12 as what Meta itself calls its first frontier model, built at the company's Prometheus data center in Ohio. Independent benchmarking from Artificial Analysis puts Wonder Full at 56 and Wonder at 52 on its Intelligence Index — genuinely competitive numbers, and a more concrete independent data point than anything OpenAI or xAI generated this month. Pricing for Wonder runs $4 per million input tokens and $20 per million output tokens via the Meta Model API and Muse Code; Wonder Full is Muse Code-subscribers-only, with no API access and no open weights for either model, a real departure from Meta's earlier Llama-era openness that the announcement doesn't explain.

That launch landed alongside Meta's first public growth-slowdown story. Hunterbrook Media's TickerTrends report, published October 9, reported Muse's daily active user growth rate slowing roughly 62% from its early-September surge, alongside a reported download decline — but Hunterbrook Capital held a short position in Meta stock at the time of publication, a disclosure this series treats as mandatory to state alongside any figure drawn from that report, following the detailed look at that conflict here. A separate data point from Sensor Tower, a firm with no disclosed short position, put US downloads at more than 5 million in Muse's first 22 days — a different metric (cumulative, US-only) measuring something distinct from Hunterbrook's daily-rate figures, not a direct contradiction of them, as this series lays out explicitly.

Three companies, three different risk profiles this month: OpenAI moved carefully and said little about its newest product specifically; xAI moved boldly into genuinely new territory with the least documentation; Meta shipped the most independently verifiable technical win of the three while absorbing the most public scrutiny of its growth numbers.

It's worth naming the underlying strategic logic behind each company's choice this month, because it's not random. OpenAI, already the incumbent with the largest existing consumer user base through ChatGPT, has the least incentive to move fast and loud on a brand-new product category — steady iteration and letting the grace period quietly expire is a defensible way to manage a feature that's still finding its footing without inviting extra scrutiny. xAI, the smallest and newest entrant of the three in terms of consumer reach, has the most incentive to make a dramatic, attention-grabbing move, and handing Grokipedia's pipeline to Grok Bots is exactly that kind of move — bold, public, and hard to ignore. Meta sits in between: large enough to afford a genuine frontier-model investment like Wonder, but exposed enough to consumer-trust scrutiny, given its own history with past products, that a growth-slowdown report lands as a real reputational risk rather than background noise.

The one thing none of them did

Every one of these developments shares the same gap this series flagged for all three products at the thirty-article mark in September: none of the three companies has published a full safety or red-team report for its newest capability specifically — not OpenAI's post-grace-period terms for dots, not xAI's editorial safeguards for a Grokipedia pipeline now run by Grok Bots, not Meta's promised-but-not-yet-published safety report for Muse Wonder. A month of real, often substantial product news across all three companies, and the specific documentation this series has asked for since its first thirty articles remains unpublished across the board.

Part 46 of 47 in the Grok Bot series. Previously: Grok Bot's Safety Framework, Still Missing, Still Not Nothing. Next: Grok Bot's First Ten Weeks: Beta, Business and Now an Encyclopedia. Browse more builder insights or explore AI skills at aiskill.market.