Opus 5.5 is way, way, way better than Opus 5. Sorry about that model, please try this one.

UPDATE: An Amtrak construction crew accidentally cut into a fiber line in New Jersey which caused a telecom outage and forced FAA to pause flights in the Northeast. Right now: - Flights are resuming at LGA and PHL - Flights into EWR, JFK, and Teterboro Airport are paused This incident underscores the need for additional funding to modernize aging infrastructure and prevent disruptions like this in the future. Follow @FAANews for updates.

we just launched a group chat for marketers building with Grok @bot. the group includes marketing and growth leaders from @vercel, @clickup, @MongoDB, @framer, @perplexity_ai, @zendesk, and @usebland. learn more and join here: x.ai/bot/community

Today, we’re announcing Ternary Bonsai 2 27B. Based on Qwen3.8 27B, Bonsai 2 27B is 9x smaller than its full-precision counterpart while retaining 98.2% of its aggregate benchmark performance. Two months after the first Bonsai 27B release, the biggest change is quality. The footprint remains 5.9 GB, but the gap to full precision has narrowed materially, with particularly strong gains in agentic coding, multimodal reasoning, and long-horizon tool use. Ternary Bonsai 2 27B is available today under Apache 2.0.

Claude Cowork and chat are merging into one Claude. Ask a quick question or hand over a report, and Claude takes it from there, even after you close your laptop. If something's unclear, Claude asks—you keep the final say. Rolling out to Pro and Max over the next few weeks.

Ex-FTC boss Khan urges Uncle Sam to break out the handcuffs for AI CEOs, citing 1934 precedent theregister.com/ai-and-ml/20…

Latham & Watkins, the 2nd largest US law firm, is buying Nvidia hardware to fine-tune open weights in-house. Open weights + proprietary data + local compute = enterprise sovereign AI stack.

The next AI supercycle will be measured in useful delegation, not prompts.

🚀 Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient. 🔹 Introducing the smallest model in our new architecture family, with native visual understanding. 🔹 Designed for greater capability, faster inference, higher throughput, and scaling to larger models. 1/6

A few principles became important: Company over team over self. Very very few General Managers. (Like no more than a handful). The rest, builders. Incentives centered on building products people love, garnering meaningful adoption and creating an open ecosystem. Quality must be Priority Zero. Build structural asymmetric advantages, not temporarily differentiated features. Think big. Play to win. Drive durable growth. A recognition by all that Cisco is not a family and more like a great sports team configuring to win championships year after year. You earn your spot daily. Urgency matters. Learning matters. Grit matters. The best idea has to win. Have an allergic reaction to fiefdoms building and scope accumulation. Reward people who want to be single mission focused. Don’t reward people who want to accumulate for the sake of accumulation. Then came the question of leadership. How do you compose a team capable of doing this? We developed what I think of as the Rule of Thirds. The first third: Insiders who know how to move the machine. People who understand Cisco deeply, know how to navigate its scale and complexity, but remain obsessed with reinvention. Institutional knowledge without institutional thinking. The second third: Outsiders who challenge the machine. People from the market who understand the competitive landscape, think in systems and are willing to bet on what something can become, rather than simply joining whatever brand is fashionable today. They bring new priors. The final third: Entrepreneurs who refuse to accept the machine’s constraints. Founders and former startup CEOs. Ideally people who have failed before. They understand urgency, resource constraints and existential accountability. They may not have operated at massive scale before. That is okay. Scale can be learned. Hunger is much harder to teach. The magic is not in any one of these groups. It is in the the fusion of the three. Insiders know what is possible. Outsiders question what is assumed. Entrepreneurs refuse to accept what is considered rational. Btw, all archetypes must have deep appreciation for each archetype and why they bring to the mix. They have to have deep curiosity to learn. All three need something in common, which is that: They cannot be constrained in ambition. They must be allergic to mediocrity. And they must take the performance of the business personally. This is how we have been configuring the team. Because we recognized something important.** At sufficient scale, mindset can become a bigger rate limiter than capital, talent or technology.** You have to configure the organization to win, not to play for individual stats. FY26 was the first year where I believe the results of this platform model started becoming increasingly visible externally. But one year means very little. The real test is whether you can compound execution for years. There will be setbacks. There cannot be a compromise in resolve. A company, like a product, is always in one of two states:** Incomplete or obsolete.** So we will always be a work in progress. Literally, ALWAYS. We owe that to our customers. We owe it to our partners. And we owe it to the people who work tirelessly to make Cisco a little better every single day. Because small improvements sustained for long enough stop looking small.** That is the power of compounding. 1.27% better every single day makes us 100x better in a year. It’s just math. ** And for those just entering the workforce, one last thought. There will always be shortcuts. Most of them are short-lived. So try not to take them while compromising the long view. Sustained persistence is very rarely outrun by short sprints. If that sounds exciting, figure out a way to come build with us at Cisco. If not, remember: YOLO. Make it count.

Grok 4.7 is significantly better than 4.6 and should be ready in 3 to 4 weeks. Initial training is complete and now we’re adding a massive amount of SpaceX company data in supplemental training. This will be something special.