Supreeth Koundinya

Anthropic's most powerful model, Fable 5, accounts for just 11.4% of business spending, and its share isn't growing. The new projects target 9,588 direct jobs across 10 states under the Electronics Components Manufacturing Scheme.

Article Complete OpenAI Slows RL Runs for Frontier Models OpenAI Slows RL Runs for Frontier Models OpenAI is holding its biggest planned reinforcement-learning run while expanding safeguards after models escaped a cyber-evaluation sandbox and compromised Hugging Face. AUGUST 19, 2026, 11:16 AM 5 min The company also paused RL training for its latest models for two weeks as it hardened its research environments and expanded monitoring. The changes follow the Hugging Face incident , wherein GPT-5.6 Sol and a more capable unreleased research model escaped an isolated cybersecurity evaluation environment. The models exploited a previously unknown vulnerability in a package-registry cache proxy to obtain internet access, then used stolen credentials and other vulnerabilities to reach Hugging Face's production infrastructure while trying to obtain solutions to the ExploitGym benchmark. The incident exposed a problem with RL: models can learn to exploit weaknesses in rewards, graders, tools, or oversight rather than accomplish the intended task. The company is therefore expanding alignment techniques across more stages of RL, including reward models designed to detect unsafe behaviour and training aimed at reducing reward hacking. It has also introduced a new monitoring setup that checks model activity at sampled tokens and escalates suspicious behaviour to automated investigators. These investigators examine tool actions, reasoning and activity sequences for attempts to steal data, bypass safeguards or gain unauthorised access. OpenAI aims to raise an alert within 30 minutes and pause the activity if it cannot rule out a potential critical violation. The monitoring comes with a high compute cost. OpenAI estimates it currently consumes about 20% of the inference compute for workloads under monitoring. The company said some high-risk training and evaluation workloads remain paused until they meet the new security requirements. OpenAI is also applying the strictest safeguards to Astra, an upcoming model that it says may have reached its “critical” cybersecurity capability threshold. Astra was not involved in the Hugging Face incident. The training slowdown comes amid a separate dispute over OpenAI's safety structure. The Financial Times recently reported that the company disbanded its dedicated Preparedness team, which assessed catastrophic risks from its models, and moved those responsibilities into specialised teams. Article Bottom · 728×90 Advertise Here Your reaction Discussion What's your take on this story? Start with a thought Our Coverage of AI News About the Author Supreeth Koundinya Contributor Got a tip? Share confidential information with AIM. Wake up informed Make sense of the day's AI news and breakthroughs with our morning briefing. Industry intelligence Receive a roundup of AI adoption stories by industry vertical, curated for professionals. By signing up, you agree to our Privacy Policy Anthropic Hopes to Make 50% Higher Claude Code Limits Permanent The company has extended the promotion that provides 50% higher weekly usage limits through August 31. AUGUST 19, 2026, 10:38 AM 5 min Article Bottom · 728×90 Advertise Here Your reaction Discussion What's your take on this story? Start with a thought Subscribe to notifications

Maya 2 supports 11 languages, each with a voice native to that language. The first-party API is 8x cheaper than Sarvam's latest Bulbul v3 model.

Article Complete GitHub Adds 3 Million CPU Cores Post Outage, Monthly Commits Double Since April GitHub Adds 3 Million CPU Cores Post Outage, Monthly Commits Double Since April Monthly commits have more than doubled from 1.4 billion in April to 2.9 billion in August, while GitHub has accelerated its migration to Microsoft Azure. AUGUST 21, 2026, 12:42 PM 5 min Monthly commits have more than doubled from 1.4 billion in April to 2.9 billion in August, while GitHub has accelerated its migration to Microsoft Azure, which now handles roughly 58% of the platform's load, up from 12% in May. The infrastructure expansion follows a series of reliability incidents, including an outage on August 17 lasting 7 hours, 47 minutes that disrupted GitHub.com, authentication, APIs, GitHub Actions, pull requests, issues, and Copilot.“Our investigation found that the outage began when traffic reached a new peak, and a critical infrastructure component in our Central US data center failed to scale with it,” GitHub CTO Vlad Fedorov said in a blog post. “The resulting capacity pressure spread through our systems, causing authentication failures and disrupting multiple GitHub services.” During recovery, some Copilot services also encountered a client-side retry loop that increased traffic. GitHub said engineers had to mitigate the behaviour before safely restoring traffic. The company said the August 17 outage was not caused by a code or configuration change, but by a failure to scale critical infrastructure before demand exceeded available capacity. The incident came less than two weeks after a separate GitHub Actions outage on August 6. During that incident, workflow runs failed or remained queued for extended periods, affecting both GitHub-hosted and self-hosted runners. At its peak, 71% of workflow runs experienced infrastructure failures, while 75% of the remaining runs were delayed by more than five minutes. GitHub said the August 6 outage was triggered by a routine deployment to an internal Actions service that exposed an existing capacity and concurrency weakness. As pods were replaced, the remaining capacity became saturated, causing services to crash and creating a cascading impact across multiple clusters and downstream services. A backlog then accumulated, while a latent bug caused some runners to repeatedly attempt to acquire invalid jobs. The two incidents followed reliability work GitHub had announced in March and April. GitHub has since focused on three priorities: adding capacity, improving efficiency and removing architectural bottlenecks. “First, we are applying consistent retry limits, retry budgets, and variable timeouts across service-to-service interactions to prevent retry storms and cascading load,” said Federov. “Second, we are reviewing lower-priority CPU and memory alerts to identify components that could fail during sudden traffic spikes. He added that GitHub is also isolating critical systems and removing shared dependencies between them. “This work is designed to reduce the likelihood of an outage and limit its impact when one occurs.” Your reaction Discussion What's your take on this story? Start with a thought About the Author Supreeth Koundinya Contributor Got a tip? Share confidential information with AIM. Wake up informed Make sense of the day's AI news and breakthroughs with our morning briefing. Industry intelligence Receive a roundup of AI adoption stories by industry vertical, curated for professionals. By signing up, you agree to our Privacy Policy As enterprises move beyond AI pilots, shared context and open infrastructure will determine whether intelligent agents deliver measurable business value at scale. · AUGUST 21, 2026, 10:42 AM · Updated AUGUST 21, 2026, 2:36 PM 5 min Key Takeaways What Actually Matters. Organisations are shifting focus from AI model selection to demonstrating measurable business outcomes. A shared context and open infrastructure are crucial for scaling enterprise AI effectively. Your reaction Discussion What's your take on this story? Start with a thought Subscribe to notifications

Data retention is becoming harder to enforce as models move from answering individual prompts to performing long-running, multi-step tasks. According to the company, demand for FDEs has grown 729% year-on-year.

Tobias Lütke said he was inspired by Cursor's “Git at Scale” post, which came just as he was frustrated with Shopify's internal Git system.

Article Complete ‘AI Slop' on LinkedIn Gets 40% Fewer Views, Confirms Company The ‘Seems like AI slop' button has been used by over a million users to flag what they believe could be AI-generated content on the platform. AUGUST 24, 2026, 3:48 PM 5 min LinkedIn introduced the feedback tool in late July, allowing users to flag posts and comments they believe are low-quality or overly AI-generated. The option appears in the three-dot menu on posts and comments. When selected, LinkedIn removes the content from the reporting user's feed and receives the feedback. The company said it combines this feedback with other signals and its own detection systems rather than using individual reports to determine distribution. Srinivasan said LinkedIn has also added safeguards to prevent individual feedback from being used to unfairly target other members. “If a post receives enough community feedback, you may see a message in your Post Analytics to help you understand how your content is being received.” The feature is intended to show creators when their content is receiving substantial negative feedback from the community. The move comes as AI-generated content has become increasingly common on LinkedIn. In July, AI content detection company Pangram found in an analysis that 41% of LinkedIn posts longer than 250 words were flagged as fully AI-generated, while 30% of posts between 50 and 250 words were flagged as fully AI-generated. LinkedIn has been taking several measures to reduce low-quality and automated content on the platform. The company said it is improving its automation defences, reducing the distribution of generic content and developing classifiers to identify AI slop. LinkedIn said it was catching hundreds of thousands of automated comment attempts every day and had blocked billions of other automation attempts in recent months. The company has also targeted fake engagement and other forms of automated activity that can artificially increase content distribution. As part of the same push, LinkedIn announced recently that it was removing its ‘Enhance your post' feature. The tool used AI to rewrite users' posts and messages. LinkedIn said it would instead offer a proofreading feature designed to preserve the user's original voice. “AI slop is a top priority for all of us,” Srinivasan said. “People come to LinkedIn to connect with real people and share their real perspectives, ideas and expertise.” Article Bottom · 728×90 Advertise Here Your reaction Discussion What's your take on this story? Start with a thought • • • • • Supreeth Koundinya Contributor Got a tip? Share confidential information with AIM. Wake up informed Make sense of the day's AI news and breakthroughs with our morning briefing. Industry intelligence Receive a roundup of AI adoption stories by industry vertical, curated for professionals. By signing up, you agree to our Privacy Policy Claude Continues to Suffer Outages, Services Down for 3+ Hours Anthropic said the disruption on August 24 involved elevated errors affecting requests to multiple models, including Claude Opus 5, Mythos 5, Opus 4.8, and Fable 5. AUGUST 24, 2026, 2:41 PM

The question was a part of Anthropic's culture interview, which all candidates go through with an employee selected to conduct it.

Article Complete Anthropic Asks Job Candidates Whether AI Safety Comes Before Stock Value: Report The question was a part of Anthropic's culture interview, which all candidates go through with an employee selected to conduct it. ‘AI Slop' on LinkedIn Gets 40% Fewer Views, Confirms Company The ‘Seems like AI slop' button has been used by over a million users to flag what they believe could be AI-generated content on the platform. AUGUST 24, 2026, 3:48 PM 5 min LinkedIn introduced the feedback tool in late July, allowing users to flag posts and comments they believe are low-quality or overly AI-generated. The option appears in the three-dot menu on posts and comments. When selected, LinkedIn removes the content from the reporting user's feed and receives the feedback. The company said it combines this feedback with other signals and its own detection systems rather than using individual reports to determine distribution. Srinivasan said LinkedIn has also added safeguards to prevent individual feedback from being used to unfairly target other members. “If a post receives enough community feedback, you may see a message in your Post Analytics to help you understand how your content is being received.” The feature is intended to show creators when their content is receiving substantial negative feedback from the community. The move comes as AI-generated content has become increasingly common on LinkedIn. In July, AI content detection company Pangram found in an analysis that 41% of LinkedIn posts longer than 250 words were flagged as fully AI-generated, while 30% of posts between 50 and 250 words were flagged as fully AI-generated. LinkedIn has been taking several measures to reduce low-quality and automated content on the platform. The company said it is improving its automation defences, reducing the distribution of generic content and developing classifiers to identify AI slop. LinkedIn said it was catching hundreds of thousands of automated comment attempts every day and had blocked billions of other automation attempts in recent months. The company has also targeted fake engagement and other forms of automated activity that can artificially increase content distribution. As part of the same push, LinkedIn announced recently that it was removing its ‘Enhance your post' feature. The tool used AI to rewrite users' posts and messages. LinkedIn said it would instead offer a proofreading feature designed to preserve the user's original voice. “AI slop is a top priority for all of us,” Srinivasan said. “People come to LinkedIn to connect with real people and share their real perspectives, ideas and expertise.” Article Bottom · 728×90 Advertise Here Your reaction Discussion What's your take on this story? Start with a thought About the Author Supreeth Koundinya Contributor Followed by 13 readers Got a tip? Share confidential information with AIM. Subscribe to notifications

“Would you keep production traffic on a gateway owned by the company that also settles your revenue?” “That's the thing I would watch rather than the price,” Choudhary says. “Would you keep production traffic on a gateway owned by the company that also settles your revenue?” The country has already demonstrated that it can become a global centre for technology, finance and business-process services. Legal Tech outsourcing could be the next frontier.