Falcon 9 Outpaces Space Shuttle as Grok 4.7 Tops Cyber Index

Key Takeaways

1. Falcon 9 flew 165 missions in 2025, exceeding the Space Shuttle's 135 missions across 1981–2011, while booster B1067 reached 37 launches and landings.
2. Grok 4.7 ranked first on Artificial Analysis' Cyber Index, ahead of Claude Opus 5.5, Fable 5.1, and ChatGPT 6 Astra; Grok Imagine added keyframe video generation.
3. Perplexity's Decisions API completed a Pokémon FireRed run with a 592 ms median response time and an estimated $0.028 inference cost across 137 calls.
4. Supabase raised $150M at a $10.65B valuation, with over 13 million developers and more than 1 million new users added monthly.
5. Tesla's Texas-authorized Cybercab fleet grew from 45 to 169 vehicles in one month, reaching 589 total Robotaxi vehicles including 420 Model Ys.
6. Basecamp users reported smooth collaboration with @bot and @claudeai agents that self-assign tasks and are managed through a mission control view.
7. OpenRouter's Alex Atallah and Replit's Amjad Masad debated agent specialization, model lock-in, and whether payment and inference will converge.

1. Space and Satellite Infrastructure

1. Falcon 9's 2025 launch cadence surpassed the entire Space Shuttle program: 165 missions in one year versus 135 Shuttle missions from 1981 to 2011, with Falcon 9 also exceeding all other orbital launch vehicles combined that year. At the Shuttle's historical average pace, reaching 165 launches would have taken roughly 37 years. Reuse figures reinforce the gap: Shuttle Discovery flew 39 times over about 27 years, while booster B1067 first flew on June 3, 2021, and had completed 37 successful launches and landings by August 25, 2026. 1
2. Starship is deploying Starlink V3 satellites, while the next-generation Starlink V5 uses a smaller, lighter design with improved efficiency and speeds of 375+ Mbps for home internet uses including streaming, video calls, and gaming in select regions. United's first Starlink-equipped flight operated from Austin to Los Angeles, described as a "game changer" with expectations for Japan and Middle East routes. 1 2 3

2. AI Models, Agents, and Developer Tools

1. Grok 4.7 ranked first on Artificial Analysis' Cyber Index, surpassing Claude Opus 5.5, Fable 5.1, and ChatGPT 6 Astra; the index combines CWE-Bench-AA, DeepsecBench-AA, and CyberGym-E2E-AA evaluations. A separate account said Grok 4.7 scored highest on Frontier v4 and beat GPT-6.1 Sol, Opus 5.5, and GPT-6 Astra at every effort level. Grok Imagine added keyframe video generation for first, last, and intermediate frames, appearing in the interface without an official announcement. 1 2 3
2. Perplexity's Decisions API drove a complete Pokémon FireRed run through the Elite Four and Champion, with a 592 ms median API response time, 987 ms p95, and 96.4% of responses under one second across 137 real-time calls at an estimated $0.028 inference cost. Separately, Computer can generate embedded visualizations from a "Visualize..." prompt, including interactive educational components and animations, with a 3D cutaway of a jet engine shown as an example. 1 2
3. Basecamp users reported that @bot and @claudeai agents collaborate smoothly inside the product: content stays in one place, agents @ each other, self-assign to-dos, and are coordinated through a mission control view. SpaceXAI released a free 53-minute Grok Bot workshop covering single-responsibility bots, inbox and meeting automation, lead research, parallel task execution, and a role-to-tools-to-skills-to-feedback-to-routine configuration flow. 1 2
4. OpenRouter co-founder Alex Atallah and Replit co-founder Amjad Masad discussed AI independence and specialization: Atallah said he did not originally intend to sell to Stripe and argued payment and inference will converge, while Masad said Replit is building a layer that lets enterprises use any model and any cloud without lock-in. They disagreed on personal agents, with Masad favoring one cross-domain agent and Atallah arguing 10 specialized chiefs of staff beat one superintelligent agent. Atallah also said agent loops, connectors, memory, sandboxes, and always-on agents are becoming baseline requirements rather than differentiators. 1 2

3. Funding, Robotics, and Autonomous Vehicles

1. Supabase raised $150 million at a $10.65 billion valuation, with more than 13 million developers using it as an application backend, over 1 million new users added monthly, more than 4 million databases created monthly, and 70% of new databases created by agents or AI-driven tools. 1
2. Tesla's Texas-authorized Cybercab fleet grew from 45 to 169 vehicles in one month, nearly quadrupling, and reached 589 total authorized Robotaxi vehicles including 420 Model Ys, with unsupervised Robotaxi miles growing by double digits weekly. Elon Musk said Tesla is "extremely careful" about autonomous driving safety, comparing the goal to making Tesla the safest human-driven car, while a related account noted that 30,000 to 40,000 annual US traffic deaths receive little national attention but a single Robotaxi injury could trigger global headlines and tighter regulation. 1 2
3. Elon Musk said "Tesla FSD feels like magic." A cited account said Oppenheim Group founder Jason Oppenheim sold his Bentley for a Tesla Model Y, bought Teslas for 10 employees because of FSD, and called FSD eight times safer than an average driver; the same account said FSD has become a material driver of Tesla's North American sales rather than an anecdote. Another account said longtime Porsche enthusiast Thomas went from Tesla critic to Model Y owner and now owns three Teslas, with Model Y the world's best-selling vehicle for several consecutive years across both combustion and electric categories. 1 2

4. Policy, Adoption, and Market Signals

1. California's Farmersville City Council voted 4-1 to use Grok to draft meeting minutes, with drafts generated from meeting recordings and proofread by the city clerk; the city manager called Grok more affordable, more comprehensive, and more objective. Grokipedia released v0.3. 1 2
2. A hotel booking tool added direct booking links that jump to hotel websites to bypass OTAs, or display hotel email addresses when available so travelers can contact hotels directly and save money. A cited example described a hotel listed at €160 per night with breakfast on Booking; after a text message noting high platform commissions and offering cash direct booking, the hotel replied with a €90 rate. 1
3. Paul Graham argued that the biggest misconception about startups is that founders start with a brilliant idea and then implement it, when larger ideas often emerge later; he said he was talking with a YC founder who had been building for 15 years and was still doing major new things. He also rejected claims that AI cannot conjecture, reason, or explain, saying those descriptions fit last spring's models and will not apply to anyone within months. 1 2
4. Guillermo Rauch said AI will make everything free, including AI itself, with free energy as the final domino and humanity's ultimate frontier. He argued security is becoming an increasingly important function at software companies, serving both as verification engineering and capital allocation, and said small teams still have large areas to disrupt given poor global cybersecurity and growing centralization, expressing investment interest. 1 2