JE Labs(@JELabs2024):Model Watch #1: GPT-5.6 (Sol), for marketers We moved our whole team to GPT-5.6 Sol this week. Here's the 7-day verdict, and why it matters more to marketers than to engineers. 🧵 What Sol is? The short version: GPT-5.6 Sol (@OpenAI) finishes exactly what you spec. Give it a narrow instruction and it does that much, no detours. It's also cheaper than Fable and burns fewer tokens. Predictable means easy to trust and delegate. The one catch: hand it open-ended work and it over-reviews. Verifies before fixing, verifies after, then finds something else to inspect. The thoroughness is real, but the loop stalls actual progress. The fix is a workflow, not a prompt: • Analysis-only session first: "don't touch code, just report." • Split problems into parallel sessions • Fix sequentially, using each report as the spec Now the part we actually care about as an agency. Same shift, four marketing reads 👇 1/ Conversion comes from hands-on reviews, not spec sheets. What moved GPT-5.6 was "I used it for days" content (Lenny's, Every's Vibe Check). In KOL briefs and PR, lead with how someone's workflow changed, and let the score table sit in the appendix. 2/"Humans frame, AI executes" is your content workflow too. You own strategy, angle, tone. AI does the reps: variations, research synthesis, first drafts. Don't pile it into one session. Separate research, drafting, and polishing so it can't wander off. 3/Cheap tokens mean volume is cheap. Sol runs ~1/3 to 1/2 of Fable, output tokens down ~54%. Anyone can now pump channel variations and multilingual versions. So the differentiator moves back to the human-owned part: the angle and the insight. 4/A user-growth surge is a timing signal. Codex's single-day user growth after launch beat the prior two weeks combined. When attention floods a category, ride it: hands-on content, ambassadors, comparison reviews. "The model got better" is your best window. And build a stack, not a hero tool. We split Opus for UI, Sol for bugs. Do the same with your marketing tools: right tool per stage, humans connecting the seams.

JE Labs@JELabs2024

2026.07.14 11:14

Model Watch #1#: GPT-5.6 (Sol), for marketers We moved our whole team to GPT-5.6 Sol this week. Here's the 7-day verdict, and why it matters more to marketers than to engineers. 🧵 What Sol is? The short version: GPT-5.6 Sol (@OpenAI) finishes exactly what you spec. Give it a narrow instruction and it does that much, no detours. It's also cheaper than Fable and burns fewer tokens. Predictable means easy to trust and delegate. The one catch: hand it open-ended work and it over-reviews. Verifies before fixing, verifies after, then finds something else to inspect. The thoroughness is real, but the loop stalls actual progress. The fix is a workflow, not a prompt: • Analysis-only session first: "don't touch code, just report." • Split problems into parallel sessions • Fix sequentially, using each report as the spec Now the part we actually care about as an agency. Same shift, four marketing reads 👇 1/ Conversion comes from hands-on reviews, not spec sheets. What moved GPT-5.6 was "I used it for days" content (Lenny's, Every's Vibe Check). In KOL briefs and PR, lead with how someone's workflow changed, and let the score table sit in the appendix. 2/"Humans frame, AI executes" is your content workflow too. You own strategy, angle, tone. AI does the reps: variations, research synthesis, first drafts. Don't pile it into one session. Separate research, drafting, and polishing so it can't wander off. 3/Cheap tokens mean volume is cheap. Sol runs ~1/3 to 1/2 of Fable, output tokens down ~54%. Anyone can now pump channel variations and multilingual versions. So the differentiator moves back to the human-owned part: the angle and the insight. 4/A user-growth surge is a timing signal. Codex's single-day user growth after launch beat the prior two weeks combined. When attention floods a category, ride it: hands-on content, ambassadors, comparison reviews. "The model got better" is your best window. And build a stack, not a hero tool. We split Opus for UI, Sol for bugs. Do the same with your marketing tools: right tool per stage, humans connecting the seams.