WEBVTT 00:00:00.000 --> 00:00:04.140 Already using Codex, Claude, or another coding agent? 00:00:04.140 --> 00:00:05.880 MOMM gives it a review team. 00:00:05.880 --> 00:00:09.860 Instead of copying work between separate chats, make one request. 00:00:09.860 --> 00:00:13.940 Your selected AI tools challenge the work, and your agent brings the 00:00:13.940 --> 00:00:14.960 result back together. 00:00:14.960 --> 00:00:16.260 Keep your agent. 00:00:16.260 --> 00:00:17.780 Add the review team. 00:00:17.780 --> 00:00:20.500 MOMM stands for Mixture of Model Modality. 00:00:20.500 --> 00:00:24.364 Think different AI reviewers, working with different kinds of input: 00:00:24.364 --> 00:00:28.614 code, written plans, and, where a reviewer supports it, images or 00:00:28.614 --> 00:00:29.000 documents. 00:00:29.000 --> 00:00:31.740 It's not a new model. 00:00:31.740 --> 00:00:34.080 It's a way to make your existing tools work together. 00:00:34.080 --> 00:00:36.220 The benefit is broader scrutiny. 00:00:36.220 --> 00:00:40.494 Ask other tools to challenge an assumption, spot an overlooked risk, 00:00:40.494 --> 00:00:44.380 or suggest a simpler approach before you commit to it. 00:00:44.380 --> 00:00:48.340 Your current agent remains responsible for investigating and testing 00:00:48.340 --> 00:00:49.220 the advice. 00:00:49.220 --> 00:00:53.111 You get more challenge, without handing your project to competing 00:00:53.111 --> 00:00:53.500 editors. 00:00:53.500 --> 00:00:57.740 If you're manually copying work between chats, you wait for answers, 00:00:57.740 --> 00:01:01.980 compare the replies, then explain the changes back to your agent. 00:01:01.980 --> 00:01:05.738 MOMM runs selected reviews in parallel and collects them into one 00:01:05.738 --> 00:01:06.080 report. 00:01:06.080 --> 00:01:08.580 That's less manual coordination. 00:01:08.580 --> 00:01:13.163 It can reduce review waiting, although more reviewers still consume 00:01:13.163 --> 00:01:14.080 more quota. 00:01:14.080 --> 00:01:17.620 On September fourth, the public development record held three 00:01:17.620 --> 00:01:21.947 hundred and two apply decisions, and two hundred and fifty one 00:01:21.947 --> 00:01:22.340 rejections. 00:01:22.340 --> 00:01:24.680 Advice was weighed, not just accepted. 00:01:24.680 --> 00:01:28.740 These are the maintainer's records, not a customer benchmark. 00:01:28.740 --> 00:01:32.473 No universal productivity or accuracy percentage has been 00:01:32.473 --> 00:01:32.940 established. 00:01:32.940 --> 00:01:33.740 Why MOMM? 00:01:33.740 --> 00:01:35.680 You choose the provider tools. 00:01:35.680 --> 00:01:38.380 Keep the agent you already work in. 00:01:38.380 --> 00:01:42.480 Inspect the open source protocol and your local decision history. 00:01:42.480 --> 00:01:46.380 CodeRabbit and Qodo also offer AI review workflows. 00:01:46.380 --> 00:01:50.565 MOMM is a fit when you want to coordinate your selected account based 00:01:50.565 --> 00:01:53.140 tools, with one responsible agent and explicit verification. 00:01:53.140 --> 00:01:54.640 Here's how to start. 00:01:54.640 --> 00:01:58.760 You need Git, Node eighteen or newer, and a compatible agent. 00:01:58.760 --> 00:02:00.880 Keep the clone in a permanent folder. 00:02:00.880 --> 00:02:04.758 Open the MOMM guide, copy the installation and upgrade prompt, and 00:02:04.758 --> 00:02:06.520 paste it into your agent. 00:02:06.520 --> 00:02:10.575 It finds the stable release and previews installation for your 00:02:10.575 --> 00:02:10.980 harness. 00:02:10.980 --> 00:02:12.140 Approve the plan. 00:02:12.140 --> 00:02:14.460 Open the local Setup Center. 00:02:14.460 --> 00:02:17.940 Choose your current agent, connect a reviewer with its account 00:02:17.940 --> 00:02:19.680 login, and verify the connection. 00:02:19.680 --> 00:02:24.208 Start with one; add another when its perspective earns the extra 00:02:24.208 --> 00:02:24.620 quota. 00:02:24.620 --> 00:02:28.670 The dashboard shows command line versions and guides explicit 00:02:28.670 --> 00:02:29.120 updates. 00:02:29.120 --> 00:02:32.220 Provider subscriptions and usage limits still apply. 00:02:32.220 --> 00:02:34.020 Return to your project. 00:02:34.020 --> 00:02:36.120 Choose which providers may receive the work. 00:02:36.120 --> 00:02:39.040 Then ask: use MOMM to review my changes. 00:02:39.040 --> 00:02:42.862 Investigate findings, verify any fixes, record every decision, and 00:02:42.862 --> 00:02:44.560 show me the ledger. 00:02:44.560 --> 00:02:47.240 Your agent handles the review loop. 00:02:47.240 --> 00:02:50.060 You don't have to relay each conversation yourself. 00:02:50.060 --> 00:02:53.300 You're seeing MOMM review this film's own script. 00:02:53.300 --> 00:02:55.300 The result isn't just a score. 00:02:55.300 --> 00:02:58.500 See each reviewer's findings and suggestions together. 00:02:58.500 --> 00:03:02.212 Your agent tests material claims, applies changes the evidence 00:03:02.212 --> 00:03:05.100 supports, or rejects advice with a reason. 00:03:05.100 --> 00:03:07.320 Unfinished decisions stay visible. 00:03:07.320 --> 00:03:09.300 Agreement doesn't replace evidence. 00:03:09.300 --> 00:03:11.100 Then open your private ledger. 00:03:11.100 --> 00:03:12.780 Who raised the issue? 00:03:12.780 --> 00:03:14.180 What did your agent decide? 00:03:14.180 --> 00:03:16.360 What evidence supports the change? 00:03:16.360 --> 00:03:20.780 The report, decisions, and completion record are there to inspect. 00:03:20.780 --> 00:03:25.380 That makes review easier to revisit, explain, and hand over. 00:03:25.380 --> 00:03:28.868 Use MOMM for a meaningful change, a difficult decision, or a release 00:03:28.868 --> 00:03:29.740 you want challenged. 00:03:29.740 --> 00:03:34.116 Start small and judge the value on your own work: useful findings, 00:03:34.116 --> 00:03:35.940 review time, and rework avoided. 00:03:35.940 --> 00:03:37.100 Open the guide. 00:03:37.100 --> 00:03:38.000 Copy the prompt. 00:03:38.000 --> 00:03:39.340 Keep your agent. 00:03:39.340 --> 00:03:40.151 Add the review team.