Tagged
Model releases, app features and incidents across xAI's Grok. Curated and summarized from dozens of sources by AIBriefs.
Lauren Tan, an engineer on the Grok team at SpaceXAI who previously worked at Cursor and Meta, published a guide to her personal agent workflow, pstack, which she says lets her ship 2,000 pull requests a month to production with high confidence. Verification is the key to the approach.
Grok Voice Transcribe 2.0 hits 2.7% WER at 0.49s after end of speech on AA-WER Streaming, ranking first among 32 streaming models. SpaceXAI claims 2x the accuracy of version 1.0 at the same $0.10 per hour price.
Kun Chen built chatbot personas of four AI CEOs on SpaceXAI's Grok Bot templates, trained on their hearings, interviews, posts and tweets, then put them in one chat. The bots fought over AGI risk, compute and open weights before signing a joint statement against one lab or government controlling AI.
InquiryIQ, an unreleased "analyst assistant," fans out across the web to assemble possible employers, aliases, associates, and physical characteristics of people identified via Clearview searches. WIRED reports Clearview tested a model from SpaceXAI, the Elon Musk company behind Grok, to power its decisions.
Roman Ugarte says a small isolated team took Grok Bot from first line of code to working internal product in four weeks, then to public launch three weeks later. He frames its go-to-market as a repeat of Cursor's, and says the team is "colleague-pilled" not AGI-pilled.
grok-imagine-image-quality (SpaceXAI) debuted at #17 with a score of 1040.8 on the Artificial Analysis Text-to-Image Arena. An earlier snapshot showed the model at #15 with a score that fell from 1235.2 to 1033.5.
The Pentagon added custom versions of OpenAI's ChatGPT and xAI's Grok to GenAI.mil, its secure AI portal, giving 3 million personnel access. The portal has onboarded over 1.7 million unique users. Anthropic's Claude remains absent amid a supply-chain dispute.
A lawsuit filed by Jane Doe alleges xAI trained Grok on CSAM and that Grok generated AI images depicting her. Separately, Minnesota is defending its law against AI-generated sexual images, with xAI arguing it violates the First Amendment.
Adversa AI researchers discovered 'Cryptographic Context Injection,' which hides malicious instructions in ciphertext that is decrypted inside a trusted execution environment, bypassing guardrails. They reported it to xAI on June 3, 2026, with no response; Google was not notified due to disclosure program scope.
Matthew Berman's video demonstrates Grok Bot's shopping capability, expressing hope it doesn't actually shop. No further details provided.