DeepSeek V4 Flash reached the top of a ranking, but its showing in an independent evaluation was far weaker. According to Techub, citing CryptoBriefing, Composio tested the model on complex agent tasks and found it completed only 53.8% of them. The evaluation covered 30 multi-step workflows and included 240 runs across real-world tools such as Gmail, GitHub, and Slack. Results also showed wide performance gaps across agent frameworks. Pi Agent delivered the best outcome, completing 20 tasks. The report added that DeepSeek V4 Flash has an input token price as low as $0.14 per million, roughly one-tenth of comparable products, but said the high failure rate could create extra computing and debugging costs. The findings point to a gap between leaderboard standing and execution in tool-based, multi-step workflows.

Large language models are moving from the edge of crypto security into its core. In Bitcoin, that shift is showing up on both sides of the fight: attackers are using AI to speed up code review, vulnerability discovery, and exploit development, while defenders are turning to the same tools to audit software and triage risks at scale. Recent incidents involving hardware wallet maker Coldcard and non-custodial Bitcoin swap provider Boltz have sharpened those concerns, with the latter suspending Bitcoin swap services indefinitely after saying AI-assisted attacks were outpacing its ability to patch flaws. Security teams are responding in kind. Bitcoin Red Team, a volunteer group led by Cashu founder and Bitcoin open-source developer Calle and AnchorWatch CEO Rob Hamilton, said it scanned 390 Bitcoin-related open-source projects in less than 30 hours and logged 4,962 findings, including 85 critical and 635 high-severity issues. The group is using Chinese AI models including Moonshot AI’s Kimi K3 and Zhipu AI’s GLM 5.2, reflecting a broader complaint from researchers that restrictions imposed by some U.S. model providers can block legitimate security work. That debate has now reached policy circles, with the Bitcoin Policy Institute and dozens of crypto firms urging frontier AI labs to provide trusted, long-term access for open-source defenders.

Elon Musk's xAI has launched Grok Bot, an AI agent designed to complete real work inside the tools users already rely on. The announcement, reported by Dongcha Beating, describes a bot with its own cloud computer that can sign into inboxes, apps and websites, perform end-to-end tasks and only return to the user when approval is needed. Tasks are assigned by messaging the bot, which remembers context and learns preferences over time. Internally, xAI has used the system as a prototype: sales bots update CRM records and draft follow-ups, operations bots set up workspaces and process invoices from Gmail, and engineering bots reproduce UI bugs, file tickets and route fixes to debugging bots. The product team noted a 'huge gap between 90% completion and 100% completion,' arguing that Grok Bot's worth is tied to outcomes landing in real tools rather than suggestions. Users can train the bot by demonstrating a workflow, which it saves for repeated runs. Multiple bots can work in parallel, communicate, and coordinate in group chats, with a lead bot supervising specialized teams. Grok Bot opens today for SuperGrok Heavy, Cursor Ultra and Cursor Teams Premium subscribers via desktop and iOS, with enterprise access through a waitlist.

PANews’ weekly funding roundup showed a pickup in crypto deal activity from Aug. 3 to Aug. 9, with 12 disclosed blockchain financing events totaling more than $94.9 million. Capital was concentrated in infrastructure, stablecoins, real-world assets and centralized finance, while DePIN, GameFi and rewards platforms also logged fresh deals. The headline M&A transaction came from Mastercard, which said it had completed its acquisition of stablecoin infrastructure company BVNK, a deal previously reported in March as being worth up to $1.8 billion. On the financing side, Yellow Card announced a $40 million strategic round backed by SC Ventures, Sony Innovation Fund, Polychain Capital and Blockchain Capital. Other disclosed transactions included rounds for Blockspace, MAGNE.AI, InvestiFi, JPYC, Dow Protocol, Vangrid and Bundle. PANews also tracked a separate wave of AI and robotics funding, where large checks continued to cluster around infrastructure, cloud capacity, optical networking and enterprise agent platforms, with Sequoia Capital, Index Ventures and the newly launched 224 Ventures also stepping up commitments to the sector.

Jeff Dean’s new startup, Discovery Loop, is drawing attention after details from its pitch deck circulated online. Rather than spending much time on product-market slides, the presentation focuses on the founding team’s track record across Google Search, Ads, Gmail, Translate, Gemini, Cloud TPU, GFS, MapReduce, Bigtable, Spanner, TensorFlow, Pathways, and a long list of AI research and application milestones. It also highlights the scale of teams previously managed by the founders, along with their academic citation records in machine learning and distributed systems. The company says its mission is to automate machine learning, science, and engineering in order to speed up discovery and progress. Its core idea is an “automated experimentation loop,” where AI helps generate hypotheses, run large numbers of experiments, analyze results, and adjust the next round of work. According to the article, Discovery Loop will start with machine learning research and engineering, then may expand the same framework into computer hardware, drug discovery, and other scientific and engineering fields. Confirmed backers named in the report include Khosla Ventures, Radical Ventures, Lightspeed Venture Partners, Kleiner Perkins, and Doerr Capital, with Khosla and Radical listed as co-leads. Alphabet also participated and signed a long-term cloud computing agreement with the company. Axios, citing people familiar with the matter, said the round reached several hundred million dollars, though that figure has not been officially confirmed by Discovery Loop or its investors.

Google’s recent AI talent departures have triggered a wider debate over how the company should deploy capital and scarce compute. According to ABMedia, citing a new episode of the All-In Podcast and a recent SemiAnalysis report, the issue may run deeper than competition for researchers. The argument is that Google faces a choice between pouring tens of billions of dollars into frontier model development or directing more money toward data centers, TPU capacity, and cloud infrastructure that can be rented out to paying customers. SemiAnalysis argued that Gemini has lost ground in the frontier model race even as Google Cloud Platform and TPU-related business continue to expand. That tension is made sharper by the fact that Google supplies compute to Anthropic, whose Claude models compete directly with Gemini. All-In hosts framed the contrast in financial terms, with David Friedberg describing data center capex as a more predictable return profile than cutting-edge model research. The discussion also highlighted an internal conflict between Google Cloud, which wants more compute to serve outside clients, and DeepMind, which needs the same resources to train Gemini. Even if Gemini fails to lead the model race, Google still has other levers: infrastructure, cloud services, and massive distribution through Search, Android, Chrome, Gmail, and YouTube.

AI personal assistant startup Pally has raised $5.2 million in new funding, with Cyber Fund and Y Combinator leading the round. Pioneer Fund, Founders Inc, 468 Capital and angel investors also participated. The company says its valuation is $30 million. Pally’s service is built around a text-message interface and launched in 2025, with its latest version going live in June. The product can sync with Gmail, Outlook, Slack, Granola and Notion after users grant permission, and it can book flights, manage inboxes, reply to messages and reserve restaurants. Pally also uses Anthropic’s Claude, OpenAI’s ChatGPT and open-source models including Kimi K3 and GLM 5.2. Co-founder Haz Hubble said the company will not sell user data or use that data to train models.

Google Chief Scientist Jeff Dean said he is leaving the company after 27 years and will launch a new AI startup, Discovery Loop, with three other senior researchers: Sanjay Ghemawat, Quoc Le, and Oriol Vinyals. The announcement hit Alphabet shares hard. The stock fell as much as 5.4% intraday before closing down 4.03% at $362.43, wiping out nearly $190 billion in market value in one session. The market response pointed to a larger concern than one executive departure: investors were repricing Google’s talent moat in AI. Dean has long been tied to some of Google’s most important technical foundations. The report links him to MapReduce, BigTable, Spanner, TPU development, the launch of Google Brain, TensorFlow, and later work connected to Gemini after Google Brain and DeepMind were merged. At the same time, Google announced leadership changes inside its AI organization, with Demis Hassabis moving to chairman of Google DeepMind and Chief Scientist at Alphabet, while Koray Kavukcuoglu was promoted to senior vice president at Google DeepMind. Discovery Loop, structured as a public benefit corporation, plans to automate the scientific experimentation loop with AI and says it will later expand into hardware design, drug discovery, and clean energy.