journal
notes fromthe studio.
- 25gpt-5.6 pricing: what luna's 5x cut does to cost-to-serveopenai's new gpt-5.6 luna tier costs 5x less than sol. here is what that spread does to your ai feature's cost-to-serve before the next finance review.
- 26ai for lead qualification: score and route inboundan ai layer that scores, enriches, and routes every inbound lead against your icp and crm, so reps only open the ones worth a call. no new sdr, no chatbot.
- 27five days to under two: a month-end close case studyan anonymized case study: a multi-entity services group cut its month-end close from 5 days to under 2 with ai-assisted reconciliation, a human signing off on every line.
- 28what claude tag in slack means for ops busyworkanthropic put a persistent @claude agent in slack on june 23. here is what it changes for an ops lead drowning in lookups and approvals, and what it doesn't.
- 29ai for sop automation: a process that survives a resignationai for sop automation extracts, validates, routes, and logs the steps of your key workflow. so the process survives when the one person who built it leaves.
- 30how an 18-agent support team deflected a third of its ticketswe shipped internal retrieval plus draft-and-review for a mid-market e-commerce cx team. ~32% of repeat tickets resolved at first touch. no chatbot.
- 31governed ai hit the headlines. what it means for compliancesnowflake and anthropic made governed ai front-page news on 2026-06-02. here is what that lowers the bar on for compliance and risk teams reviewing contracts.
- 32ai for ticket triage: route, tag, and prioritize the queueai for ticket triage sits behind your agents to classify, route, and prioritize tickets. faster first response, a sane queue, no new customer-facing chatbot.
- 33how an 8-rep sales team took reply rates from 4% to 11%a research-plus-drafting layer pre-populated each rep's outbound the morning of their call list. reply rate went 4% to 11%. deliverability held.
- 34gpt-5.5 instant is chatgpt's default: what it meansOpenAI made gpt-5.5 instant the default chatgpt model with 52.5% fewer hallucinated claims. here is how a content lead should re-tool the content pipeline.
- 35ai for product teams: from v0 to a feature that sticksthe path from a Lovable or v0 ai prototype to a production feature users adopt. what to build inside your own surfaces, with the support-ticket drop as proof.
- 366 hours to 90 minutes: a content production case studywe shipped a brief-to-draft content system on Claude + Airtable for a 12-person marketing team. throughput up 3x, time-per-post down from 6 hours to 90 minutes.