SPEAKER_00 0:00

Welcome back to Sumo Growth AI Insights. I'm walking you through this week's AI news the way an operator actually uses it, not as a pile of headlines. Lead story. Anthropic and OpenAI rolled out cheaper models after the AI slowdown call. On Tuesday, both labs each shipped new, less costly models. The first releases from either lab since Anthropic CEO Dario Amode's call to Paste the Frontier and public agreement from Sam Altman and Elon Musk. OpenAI added GPT-6 Sol and GPT-6 Luna, cutting API prices by about 50% versus GPT-5.6 promotional pricing. Soul sits a tier below Astra for complex work like coding, while Luna targets high volume extraction and summarization. Anthropic unveiled Claude Opus 5.5 in its Claude 5.5 family, a more token-efficient release that costs about 40% less to run than Opus 5 per Anthropic's Diane Pen. Both labs are racing cheaper open weight rivals and customers trying to rein in AI spend. Plus story, Trump told the UN he wants to encourage AI as superintelligence, not rein it in. At the 81st UN General Assembly on September 22nd, President Donald Trump said he sought to rebrand artificial intelligence as superintelligence and planned to encourage it, not rein it in. The line lands against the same week's product drops and against a safety push that now includes roughly 1,400 Frontier Lab employees, including anthropics Dario Amode and Jack Clark, asking the U.S. government to back deliberate pacing of frontier automated AI development. For operators, the split is the story. Labs shipping cheaper models while Washington signals acceleration, not a freeze. Three more headlines. First, OpenAI proposed U.S.-led global AI standards for alignment and recursive self-improvement. On Monday, OpenAI posted safety and security proposals focused on alignment research and recursive self-improvement, or RSI, arguing the United States should lead shared frontier standards and that RSI should not be pursued unless and until it can be done safely. The lab has also said it does not yet know how to safely reach aligned full RSI. Second, Texas Governor Abbott expanded a data center halt to environmental permits. Abbott ordered the Texas Commission on Environmental Quality to stop data center-related permits until ERCOT finishes an audit of the grid connection wait list, widening August's pause on new data center grid approvals. Bloomberg NEF analysis cited in coverage put almost 20% of the U.S. 253 gigawatt data center pipeline at risk with a Texas pause. Third, Anthropic's IPO window is still in play for late September or early October. Coverage dated September 22nd says Anthropic confidentially filed a draft form S1 in June and may go public as early as late September or early October, with no trading date announced yet. Second quarter revenue grew 142% quarter over quarter, with Bloomberg-reviewed figures above $11.5 billion in preliminary Q2 revenue. Tools of the day.tech provides agents that provision their own machines, accounts, and budgets so long-running build and ops jobs keep moving without you babysitting every step. Niceai at nice.ai offers autonomous marketing agents that research, create, schedule, and publish social content with persistent brand memory for lean agency teams. Claude at Claude.ai is Anthropic's Claude Opus 5.5 in the new Claude 5.5 family, a more token efficient Opus tier priced for heavier client workloads. Business AI tip of the day reprice your agent stack before clients ask. OpenAI and Anthropic just cut model costs while Washington talks acceleration and Texas slows data center permits. That is not a thought piece for agencies. It is a margin event. This week for every Go High Level, sumo AI CRM or client agent you run, benchmark the same prompt suite on CLARD Opus 5.5, GPT-6 SOL, and GPT-6 Luna or Astra where quality still wins. Put high volume CRM summarization, call notes, and inbox triage on the cheapest tier that holds quality. Keep coding and multi-step agents on SOL or Opus 5.5. Cap monthly token spend per subaccount and alert when a bot spikes after a model swap. Keep a one-page model card on the opportunity, provider, model ID, price tier, and fallback. Use NISE AI for client social execution only after brand memory and approval gates are set. And use Solid for long-running builds with a human review before spend or account creation. Agencies that ignore the price drop will overcharge clients or underprice themselves. Agencies that reprice this week win the next retainer conversation. Pro tip ship a 15 minute cost cut loom to each retainer client showing before and after token cost on one live workflow. Offer a shared savings credit if they expand automation volume this month. That is your Sumo Growth AI Insights briefing. Reprice the agent stack this week, and I will catch you on the next issue.