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

Three personal AI agents launched within roughly three weeks of each other in September 2026: Meta's Muse, OpenAI's Dots, and xAI's Grok Bot. The original 30-part run of this series, alongside its companion coverage of the other two products, compared them at launch on the dimensions that mattered then — distribution, permissions architecture, and who each company was betting would adopt first. A month on, the more interesting comparison isn't about launch positioning anymore. It's about what each company actually chose to do with its second month, because the three choices diverged sharply, and each one tells you something real about each company's priorities.

Meta: a frontier model upgrade

Meta's answer to "what's next" was the biggest single move of the three: on October 12, it shipped Muse Wonder and Muse Wonder Full, its first "frontier model," built on top of the lightweight Muse Spark that ran the original launch. Independent benchmarking from Artificial Analysis put Wonder Full at 56 and Wonder at 52 on its Intelligence Index — real, third-party numbers rather than Meta's own marketing claims. Wonder is priced to compete with GPT-6 Sol and Opus 5.5 on the API side, while Wonder Full stays exclusive to Muse Code subscribers. That's a genuine capability bet: Meta used its second month to make the model underneath Muse meaningfully more capable, with new behaviors — Active Memory and Active Focus — aimed specifically at the kind of sustained, multi-step work a personal agent is supposed to do. It's also the month Muse's growth numbers came under public scrutiny for the first time, via the TickerTrends download-and-DAU report covered earlier in this batch, which means Meta's biggest product news and its first real growth-skepticism story landed in the same few weeks.

xAI: a much riskier bet

xAI's second month took the opposite shape — not a new model, but a new responsibility handed to the existing one. On October 8-9, Elon Musk confirmed on X that Grok Bots would now run Grokipedia's entire content pipeline — creation, fact-checking, edit review, and curation — as a coordinated multi-agent workflow, replacing the earlier single-model process, running on the Grok 4.7 model that's been stable since late September. Grokipedia itself grew from 885,279 articles at its late-2025 launch to more than 6 million by early 2026, and as of earlier this year Grok was already handling more than 75% of edit suggestions and approvals without human review. What changed in October is the last layer of human editorial oversight going away entirely for an encyclopedia with no volunteer-editor community or talk-page consensus process to fall back on — a structurally different kind of risk than a feature update, because it concentrates correctness and bias-management into a single AI system with no built-in human backstop. xAI also shipped smaller, concrete product updates in the same window — Secure Forms for filling out pages the bot already has open, and a reconnect prompt for expired X or Vercel logins, per xAI's own changelog — alongside unverified but widely circulated growth claims putting Grok Bot's daily active users somewhere around 450,000, up from roughly 155,000 a month earlier, a number that comes from a secondary source and should be read with exactly that caveat attached.

OpenAI: the quiet month

Compared to the other two, OpenAI's second month for Dots looks almost conservative. The company's own changelog shows real engineering investment — Locked Chats rolling out broadly on October 20, expanded "Finances in ChatGPT" access on October 2, and new GPT-Live-1.1 voice options on October 15 — but none of those entries mention Dots by name, and whether they meaningfully upgrade the Dots experience is inference rather than confirmed fact. OpenAI also shipped a cheaper sibling model, GPT-6.1 Sol, though Dots itself still runs on GPT-6 Astra. The most consequential Dots-specific event of the month isn't a feature at all: the usage grace period that let the first month of dot activity run free of plan-allowance limits ends in late October, and OpenAI has not yet published what replaces it. Reporting on basic rollout questions — like whether dot creation works on mobile yet — remains inconsistent across secondary sources, which is itself a sign that OpenAI hasn't been especially forthcoming with clear, dated updates the way Meta and xAI have.

Line the three up and a pattern appears: Meta spent its second month on raw capability, xAI spent it on scope and risk, and OpenAI spent it mostly in maintenance mode while the one deadline that actually mattered — the end of its usage grace period — approached without an announced plan.

Three different theories of where the value actually comes from

Step back from the specific features and the three companies' second-month choices start to look like three different theories about where a personal agent's long-term value actually comes from. Meta's bet is that the model itself is the differentiator — Wonder's independent Artificial Analysis scores and its expert-domain positioning in law, finance, and health are a wager that raw capability, paired with the Secure VM and Sentinel architecture already in place, is what wins sustained usage once the novelty of launch week fades. xAI's bet is closer to the opposite: that distribution through an existing, high-traffic surface — Grokipedia's traffic, X's user base, the Grok Build developer audience — is the thing worth expanding aggressively, even if it means taking on real, hard-to-reverse risk by removing human review from an encyclopedia's editorial pipeline before a safety framework for that specific change has been published. OpenAI's bet, at least this month, looks like patience: let the existing Dots product season with real usage, invest engineering effort into adjacent ChatGPT features that may or may not touch dots directly, and defer the one real policy decision — what happens after the grace period — rather than rushing an answer. None of these theories has been validated or invalidated by six weeks of data. They're genuinely different strategies, which is itself useful information for watching where each product goes next.

The thing none of the three has done

Across all three products, one gap repeats: no company has published a full safety or red-team evaluation for the specific new thing it shipped this month — Dots' post-grace-period terms, Grok Bot's Grokipedia editorial safeguards, or Muse Wonder's promised safety report. That's the fairest, most consistent observation to make about all three one month on, more useful than picking a winner among three genuinely different strategic bets that won't be fully evaluable until the usage data, safety evaluations, and growth trajectories each company is still sitting on actually become public.

Part 45 of 46 in the Meta Muse series. Previously: 20 Dollars or 100 Dollars a Month: Pricing the Muse Consumer Subscription. Next: Muse's First Six Weeks: From Launch Hype to a Frontier Model Bet. Browse more builder insights or explore AI skills at aiskill.market.