5 days. That is actually all it took for Clawdbot — an open-source AI assistant that guarantees to truly do issues in your laptop, not simply chat — to go viral, implode, rebrand (twice!) and emerge as OpenClaw. Bruised however nonetheless respiration as a beloved crustacean.
In the event you blinked over the previous few days, you will have missed crypto scammers hijacking X accounts, a panicked founder unintentionally freely giving his private GitHub deal with to bots and a lobster mascot that briefly sprouted a disturbingly good-looking human face. Oh, and someplace within the chaos, the AI developer Anthropic despatched a well mannered electronic mail asking them to please, for the love of logos, change the identify.
Welcome to OpenClaw. Previously Clawdbot and briefly often called Moltbot, it is the identical AI assistant below a more recent, sturdier shell. And boy, does this lobster have lore.
What even is OpenClaw? And why do you have to care?
Here is the pitch that had tech X (the platform previously often called Twitter) dropping its thoughts: Think about an AI assistant that does not simply chat; it does stuff. Actual stuff. In your laptop. Via the apps you employ.
OpenClaw lives the place you really talk, like WhatsApp, Telegram, iMessage, Slack, Discord, Sign — you identify it. You textual content it such as you’d textual content a buddy, and it remembers your conversations from weeks in the past and might ship you proactive reminders. And for those who give it permission, it could actually automate duties, run instructions and principally act like a digital private assistant that by no means sleeps. Unlike its founder.
Created by Peter Steinberger, an Austrian developer who offered his firm PSPDFKit for round $119 million after which bought bored sufficient to construct this, OpenClaw represents what lots of people thought Siri ought to have been all alongside. Not a voice-activated get together trick, however an precise assistant that learns, remembers and will get issues accomplished. (CNET reached out to Steinberger for touch upon this story.)
OpenClaw would not require any particular {hardware} to run, although the Mac Mini appears in style. The core concept is that OpenClaw itself principally routes messages to AI firms’ servers and calls APIs, and the heavy AI work occurs on whichever LLM you choose: Claude, ChatGPT, Gemini.
Hardware solely turns into an even bigger dialog if you wish to run giant native fashions or do heavy automation. That is the place highly effective machines, just like the Mac Mini, are sometimes introduced into the dialog. However that is not a requirement.
The undertaking launched about three weeks in the past and hit 9,000 GitHub stars in 24 hours. By the point the mud settled late final week, it had rocketed previous 60,000 stars, with everybody from AI researcher Andrej Karpathy to investor (and White Home AI and crypto czar) David Sacks singing its praises. MacStories referred to as it “the way forward for private AI assistants.”
Then issues bought bizarre.
The rename that broke the web (twice)
Ostensibly, final weekend, Anthropic slid into Steinberger’s inbox to level out that “Clawd” (the assistant’s identify) and “Clawdbot” (the undertaking identify) have been possibly just a bit too just like its personal AI, Claude.
“As a trademark proprietor, we have now an obligation to guard our marks — so we reached out on to the creator of Clawdbot about this,” a consultant from Anthropic mentioned in an electronic mail assertion to CNET.
By 3:38 a.m. US Eastern Time on Tuesday, Jan. 27, Steinberger made his name: “@Moltbot it’s.”
What occurred subsequent, in response to Steinberger’s posts on X and the earlier MoltBot weblog, was like a digital heist film, besides everybody was a bot and the getaway automobiles have been social media handles.
Inside seconds — actually, seconds — automated bots sniped the @clawdbot deal with. The squatter instantly posted a crypto pockets deal with. In the meantime, in a sleep-deprived panic, Steinberger unintentionally renamed his private GitHub account as an alternative of the group’s account. Bots grabbed “steipete” earlier than he may blink. He mentioned each crises required him to name in contacts at X and GitHub to make fixes.
Then there was what the creators dubbed “the Good-looking Molty incident.” Steinberger instructed Molty (the AI) to revamp its personal icon. In a single memorable try and make the mascot look “5 years older,” the AI generated a human man’s face grafted onto a lobster physique. The web turned it right into a meme (a la Handsome Squidward) inside minutes.
Pretend profiles claiming to be “Head of Engineering at Clawdbot” shilled crypto schemes. A faux $CLAWD cryptocurrency briefly hit a $16 million market cap earlier than crashing over 90%. “Any undertaking that lists me as coin proprietor is a SCAM,” Steinberger posted on X, exasperated, to 1000’s of more and more confused followers.
To proceed the chaotic saga that has unfolded over the previous week, as of Jan. 30, the undertaking has settled on renaming Moltbot to OpenClaw, bringing in “Open” for open supply and “Claw” for its lobster heritage. The identify change is smart because of these issues. Nonetheless, the reasoning is definitely a lot less complicated: Steinberger simply did not just like the identify.
What made this AI software go viral
Strip away the chaos, and OpenClaw is genuinely spectacular.
Most AI tools are principally the identical. You open an internet site, kind a query or question, look forward to it to generate, copy the reply, paste it some other place, and so forth., and so forth. OpenClaw desires to flip that script by having the assistant inside your present conversations. You are already in WhatsApp or iMessage, so why not simply textual content it such as you’d textual content a coworker?
Do not miss any of our unbiased tech content material and lab-based evaluations. Add CNET as a most popular Google supply.
The killer options? Effectively, there are three most important issues.
For one, persistent memory. OpenClaw would not overlook the whole lot while you shut the app. It learns your preferences, tracks ongoing initiatives and really remembers that dialog you had final Tuesday.
There are additionally proactive notifications. It may well message you first when one thing issues, resembling every day briefings, deadline reminders and electronic mail triage summaries. You may get up to a textual content saying, “Listed here are your three priorities at this time,” with out having to ask the AI first.
Lastly, there’s actual automation. Relying in your setup, it could actually schedule duties, fill types, arrange information, search your electronic mail, generate reviews and management sensible dwelling gadgets. Individuals reported utilizing it for the whole lot from inbox cleanup to analysis threads that span days, and from behavior monitoring to automated weekly recaps of what they shipped. The use instances appear to maintain multiplying as a result of as soon as it is wired into your precise instruments (calendar, notes, electronic mail), it stops feeling like software program and is simply a part of your routine.
Must you really use this factor?
Time for actual speak. OpenClaw shouldn’t be a refined, enterprise-ready product with vendor help and compliance paperwork — which is something Steinberger admits. It is a fast-moving, open-source undertaking that simply survived a near-death expertise involving trademark attorneys, crypto scammers and catastrophically uncovered databases. Whew.
So, you may be questioning, by all this hoopla, whether or not OpenClaw is even one thing it’s best to really attempt. Positive, this software remembers info throughout weeks, works between apps and techniques and offers proactive notifications. But it surely’s bought tough edges. This is not a software for you for those who want one thing that “simply works” and would not have sophisticated set up steps.
And also you most likely do not wish to take this on for those who do not wish to take into consideration — and do not deeply perceive — cybersecurity.
Key safety dangers to notice
Safety specialists have raised purple flags about OpenClaw’s security because it grows in reputation. As a result of the agent is designed to run regionally and work together with emails, information and credentials, even small setup errors can have huge penalties.
In latest days, researchers spotted quite a few publicly accessible OpenClaw deployments with little or no authentication, exposing API keys, chat logs and system entry to anybody who stumbled throughout them.
A few of the most seen safety considerations have been social moderately than technical, together with faux Clawdbot/Moltbot/OpenClaw downloads and hijacked accounts used to unfold malware or scams. Whereas builders have moved rapidly to patch particular flaws, safety analysts say OpenClaw’s turbulent debut highlights a bigger situation going through AI brokers: As they grow to be extra autonomous and extra highly effective, the safety dangers additionally scale simply as quick.
Roy Akerman, head of cloud and id safety at Silverfort, an id safety platform, mentioned in an electronic mail to CNET that the chance of a software like OpenClaw is not that it is overtly malicious. What’s dangerous is that it continues to behave below a respectable human id, which may blur the traces between a consumer and the machine performing on their behalf.
“When an AI agent continues to function utilizing a human’s credentials, after the human has logged off, it turns into a hybrid id that the majority safety controls aren’t designed to acknowledge or govern,” Akerman mentioned. “Organizations should not attempt to block these instruments outright, however they do want to alter their posture, deal with autonomous brokers as identities, restrict their privileges and monitor habits repeatedly, not simply logins.”
The little lobster that molted (and stored going)
In response to Steinberger, “Molting is what lobsters do to develop.” They shed their previous shell and emerge larger: from Clawdbot to Moltbot to OpenClaw.
OpenClaw is identical software program as Clawdbot, providing the identical spectacular engineering and imaginative and prescient of what private AI assistants could possibly be. However the previous almost-120 hours pressured it to develop up quick, coping with safety vulnerabilities, battening down authentication, and studying that viral success attracts not simply customers however scammers, squatters and, sure, mental property attorneys.
Via all of this, OpenClaw remains to be standing. Discord remains to be buzzing. GitHub stars preserve climbing. And someplace in Vienna (or possibly London), Peter Steinberger might be nonetheless warding off DMs from folks asking if he is launching a crypto token. (He isn’t. Please cease asking.)
Need to attempt OpenClaw your self? Head to openclaw.ai for documentation, set up guides and, most significantly, a safety guidelines.
Simply possibly use a spare laptop computer. And undoubtedly do not identify your undertaking after anybody’s trademarked AI mannequin. Seems that issues.
