A BlockTempo walkthrough matched a widely shared X thread about GrokBot against official SpaceXAI documentation and reached a simple conclusion: there is no separate "loop" or "graph" switch in the product. In practice, what users call loops maps to Routines used with Skills, while graph refers to delegation across multiple bots based on each bot’s description field. The report also lays out who can access GrokBot, listing six eligible plans: SuperGrok Plus, SuperGrok Heavy, Cursor Pro+, Cursor Ultra, and Cursor Teams Standard and Premium. It notes that GrokBot has no free tier, though SuperGrok Plus can be tried for a limited time. Eligible accounts receive weekly usage quotas and may enable pay-as-you-go billing based on model and token costs. Beyond setup, the guide focuses on operating constraints. Bots under the same account share one cloud computer, including files, browser state and login sessions, which makes handoffs easier but removes any security boundary between bots. The article also highlights seven categories of actions that require approval by default, from sending messages and publishing content to transferring funds, changing permissions and accepting legal terms. It recommends defining approval rules before problems appear, especially for teams using scheduled routines and connector-based workflows.

Tom Lee predicts Bitcoin could reach $150,000 as institutions buy crypto stocks, calling current rally just the first wave. Former SEC and CFTC officials co-signed a letter urging risk-calibrated regulation to attract offshore perpetual futures back to the U.S. Bitfinex Securities lists 5 tokenized notes linked to Strategy and other Bitcoin reserve companies. 21 global financial institutions commit to forming a joint stablecoin company targeting 2027 launch. Other key stories: OSL Group revenue up 65.8%, Kraken parent partners with LSE to tokenize 100 UK stocks, Fed's Barr warns of rate hikes if inflation persists, Singapore MAS consults on stablecoin regulation, Thai SEC proposes allowing retail to trade overseas crypto derivatives, and more.

Lawyer Lin Shanglun argues that as generative technology drives the cost of producing text, code, and charts close to zero, the assets and services with the strongest defenses may be the ones algorithms cannot easily replicate online. In his article, he points to the continued strength in physical Pokemon card and vintage sports card auctions as a case study in scarcity, shared historical value, and offline authentication. Lin says the same logic applies to professional services and enterprise software. In his view, the most durable barriers come from three areas: strict data privacy and international cybersecurity compliance, face-to-face trust built through offline business relationships, and non-public practical knowledge accumulated by experienced professionals. Public information can be read by models, he writes, but the details that often matter most in legal, medical, and operational work are not available on the open internet. He goes on to argue that AI builders should design products in reverse, embedding these “AI-resistant” elements into the system itself. That includes expert verification workflows, private feedback loops that turn offline know-how into proprietary assets, contract-level accountability structures, and human-in-the-loop delivery models that combine automation with expert judgment.

OpenAI is set to release Astra, a model featuring a recurrent depth architecture (looped transformer) that allows the same information to pass through the same layers multiple times. ByteDance and Seed team have previously published similar research. This approach enables smaller models to approach larger model performance but raises safety concerns: some reasoning occurs in hidden states, making it harder to detect violations. OpenAI has limited the use of recurrent depth and retains an interpretable chain of thought.

OpenAI CEO Sam Altman told a group of Silicon Valley interns that a four-year college degree is no longer necessary, with two years yielding maximum benefit. He argued that AI has democratized productivity, making the traditional path of elite education and internships obsolete. Altman shared his own experience dropping out of Stanford after two years to found Loopt, and cited examples of other successful dropouts like Chris Olah (Anthropic), James Dacombe (Olix), and Alexandr Wang (Scale AI). He also revealed a controversial strategy: hiring a 50-year-old proxy to negotiate with conservative companies while focusing on product development. Altman emphasized that AI allows individuals to build a company alone in a room using tokens, and urged young entrepreneurs to cut off all escape routes and go all-in on their chosen direction.

Securitize has brought its tokenized high-yield credit fund HINC to Solana lending protocol Loopscale, allowing eligible investors to borrow USDG against their fund shares without redeeming their positions. The move adds sub-investment-grade corporate credit to an onchain lending venue where collateral has largely been limited to Treasuries, government money-market funds, and investment-grade instruments. HINC, launched on Aug. 18 across Avalanche, Ethereum, Solana, and Sui, invests mainly in high-yield corporate bonds, with additional exposure to CLO tranches, bank loans, and other higher-yield fixed-income assets. The fund carries a $100,000 minimum subscription, a 0.60% annual expense ratio, and is limited to accredited investors and qualified purchasers that pass Securitize onboarding. Loopscale, which reports $91.3 million in TVL and $55.9 million in active loans, uses a RedStone oracle feed based on the fund administrator’s daily NAV to value HINC and manage liquidation events. The listing also highlights unresolved issues around tokenized credit as collateral, including liquidity under stress, allowlist restrictions, liquidation backstops, and the practical challenge of unwinding a NAV-priced asset within a bounded number of business days.

Barret Zoph, now a vice president of research at Google DeepMind, has again switched between top AI labs after prior roles at Google Brain, OpenAI, and Thinking Machines Lab. The report traces his path from early work on neural architecture search with Quoc Le to Switch Transformer with William Fedus and Noam Shazeer, then to reinforcement learning post-training at OpenAI. It also recounts the dispute around his departure from Thinking Machines Lab, where former OpenAI CTO Mira Murati later told staff his employment had been terminated over issues tied to performance, behavior, and trust, while Zoph disputed that account and said he had already planned to leave. His return to Google comes during a period of heavy management and research turnover at DeepMind, including role changes for Demis Hassabis, Jeff Dean’s exit, and moves by Noam Shazeer and John Jumper. Data cited from Zeki Data shows DeepMind’s inflow-to-outflow ratio for research and senior engineering talent fell from about 12:1 in Q2 2023 to roughly 2:1 in Q3 2026, trailing Anthropic’s 22:1 and OpenAI’s 5.7:1.

HTX said crypto markets turned sharply higher during the week of Aug. 17-23 after a months-long slump, with Aug. 19 marking the key pivot. On that day, U.S. President Donald Trump convened crypto companies, regulators, and traditional finance representatives at the White House and publicly urged Congress to move forward with digital asset legislation. The U.S. Treasury also said it would at least double the cap on single long-end Treasury liquidity support repo operations, raising it from $2 billion to $4 billion. According to the report, the 30-year Treasury yield fell by about 9 basis points that day, risk assets strengthened, and Bitcoin rose more than 8% within 12 hours. For the full week, Bitcoin gained about 22% and broke above the $60,000-$70,000 range that had held for much of the year. Between Aug. 19 and Aug. 22, the total market capitalization of crypto assets excluding Bitcoin increased by roughly $215 billion, moving back above $1 trillion. HTX said assets across several sectors rose in tandem on its platform, with Meme and DeFi names showing the most concentrated gains. TRUMP, TUT, PEPE, and PUMP all posted strong advances, while PENDLE, AAVE, and MORPHO moved higher after Pendle expanded its PT Looping incentive program. ZEC also surged after Grayscale filed an amended registration statement tied to a planned spot ETF conversion, while XRP ranked among the stronger large-cap performers of the week.