Amazon has decided to step back from the pursuit of creating the most advanced artificial intelligence. This week, the company shut down its AGI Lab in San Francisco, a team dedicated to artificial general intelligence. The closure came amid layoffs within Amazon's AGI division and occurred just 18 months after the lab was established. This move signals Amazon’s withdrawal from a race in AI development that it never fully dominated.
SAP reported a strong 22% growth in cloud revenue, reaching €6.28 billion in Q2, surpassing analyst expectations. The company's cloud backlog, an indicator of future sales, increased by 26%, boosting investor confidence. Following these results, SAP's stock climbed over 6% in Frankfurt, signaling relief amid previous worries about AI's potential negative effects on the business. Despite some stock declines earlier, the positive cloud performance has reassured the market about SAP's growth prospects.
Intel's second-quarter 2026 revenue exceeded $16 billion, marking a 25% increase from the previous year and its fastest growth pace since late 2011. The company's adjusted earnings per share reached 42 cents, doubling Wall Street's estimate of 21 cents. While shares initially rose on the strong performance, they later declined during extended trading sessions. The surge was largely driven by a 59% increase in AI data center revenue.
For the first time in internet history, automated bots have overtaken humans in web traffic. Security firm Cloudflare reports that bot-driven requests surpassed 50% in June, now comprising close to 58% of all webpage visits. Another study by cybersecurity company HUMAN supports these findings, signaling a significant shift in the nature of web interactions.
We interviewed several cybersecurity researchers who specialize in discovering unknown vulnerabilities and creating exploitation tools. They shared insights on how the safety restrictions implemented by OpenAI and Anthropic are impacting their ability to conduct offensive security work effectively.
Leading AI companies such as Nvidia and Mistral are urging US policymakers to steer clear of sweeping restrictions on open-weight AI models. This comes amid ongoing discussions in Washington about how to respond to advancements in Chinese AI technology and concerns over alleged unauthorized model replication, known as distillation.
Anduril is reportedly in discussions to secure a new funding round that could elevate its valuation to approximately $100 billion, more than three times its valuation from last year, according to Reuters.
Cognition, an AI coding startup, has acquired Poke, an AI assistant known for its friendly, conversational texting style, in a deal valuing Poke in the low nine figures. This acquisition integrates Poke’s unique interaction approach into Cognition’s coding assistant, Devin. The move highlights a shift in AI development where the personality and manner of interaction are becoming just as crucial as the underlying technology itself.
By Sumeet Vaidya
Frontier labs and major AI providers promise revolutionary innovation, and many deliver. But recent incidents like the Anthropic policy reversal and security lapses at Hugging Face and OpenAI reveal instability in these platforms. This unpredictability poses a serious risk to organizations relying on these AI models for critical operations.
Open-source initiatives such as OpenClaw and DeepSeek offer cost-effective models with comparable quality, closing previous gaps in usability, safety, and accessibility for enterprises. This evolution forces CTOs, CIOs, and engineering leaders to ask how to maintain reliability while controlling costs amid fluctuating hyperscaler pricing.
The solution is not simple but highlights the need for systems enabling rapid swapping of AI models and adaptive collaboration between AI agents and humans using real-world data and tools. Building a robust foundation ensures organizations can stay agile without merely chasing trends.
Engineering leadership has moved away from costly token-spending strategies toward sustainable practices that prioritize team morale and infrastructure modernization. Future enterprise AI must empower human-agent cooperation with equal access to real environments and strict governance over credentials and actions.
Adopting flexible architectures that accommodate model changes, integrate emerging tools, and prevent vendor lock-in builds resilience. This approach allows organizations to evolve with AI advancements steadily and securely, embodying a flexible, adaptive culture for long-term success.
Sumeet Vaidya, CEO and co-founder of Crafting, previously shaped AI engineering at Meta, Uber, and Discord.
Alphabet, Google's parent company, posted impressive second-quarter earnings with $120 billion in revenue, marking its 12th consecutive quarter of strong financial performance. These results surpassed expectations despite delays in launching the advanced AI model Gemini Pro. Investors are closely watching how Alphabet’s significant AI investments impact its financials as the company increases its annual capital expenditure forecast to $200 billion, with quarterly spending doubling year-over-year.
