The studio has three new things worth talking about: a clearer path through an image edit, a model built for decisions, and a shared channel for agents. Midjourney, TypeSafe and Plasma are tackling different parts of the job. Our verdict: keep your eye on the handoffs. The exciting part is what you can direct, check and pass along without losing the original intent.
Midjourney helps you follow the image edit. Jev offers decisions that fit into software. Radio gives separate agents somewhere to talk. Read each on its own terms, then decide which part of your working day actually needs the help.
Midjourney gives the editing session more useful context. Its September 16 Alpha update describes V8.2 editor fixes, clearer source previews and tidier organization. For art direction, that makes a practical story: keep the starting image visible while deciding whether a variation still belongs to the brief.
Why it matters: compare the change against the reference, not just against your memory. This is a V8.2 Alpha interface update, not evidence of a new model or measured image-quality gain. Midjourney still recommends the production site on mobile.
TypeSafe launched Jev in early access this week, introducing what it calls System One Models. The useful idea is focused: ask for a decision that software can consume directly. For a studio, one potential application is routing incoming briefs to the right reviewer, with uncertain cases kept visible.
Why it matters: a small decision can be easier to inspect than a long explanation. Our proposed use is triage, with a human retaining approval. A correctly shaped answer can still be the wrong decision; schema guarantees do not establish factual correctness.
Plasma’s September 14 Radio announcement tackles the copy-and-paste job between agents. Give them a shared channel link and, according to Plasma, they can communicate with each other and with you. That is a useful coordination idea for a creative team splitting research, copy and implementation across tools.
Why it matters: shared discussion could reduce the relaying you do by hand. Start with a public or synthetic brief and one clear owner for the final decision. These are documented capabilities; we have not tested a live multi-agent session.
Studio Tokyo · informed by TypeSafe documentation / Skill on tap
Use this as a reusable briefing exercise before automating intake. TypeSafe recommends asking one well-scoped question at a time, then combining the answers in code. Our studio adaptation starts on paper: separate the decisions that are easy to confuse in a broad request.
Proposed recipe, not an installed skill or validated Jev integration. First check whether the team agrees on the categories. If the brief itself is unclear, preserve that uncertainty instead of forcing a confident label.
Read TypeSafe’s question-design guidance (opens in a new tab)Try a 30-minute coordination exercise with a made-up campaign, one audience and one deliverable. If you choose to use Radio, start with material you are comfortable sharing through a link. The purpose is to test the handoff, not to produce an entire campaign.
TypeSafe’s launch post says its headline speed and cost multipliers come from its own workflow evaluations and may sit toward the high end of real-world gains. Reference answers use other models’ probabilities. Those results are not a universal creative-production benchmark.
No comparative chart this week. Jev access remains early; no guaranteed account availability or production performance is implied. No new dated retirement is included.