Praj saw the Sep 25 episode open with Jev making decisions fast and cheap, and he uses it the same way: his main model plans the work and reviews every diff, and a model he is already paying for does the bulk typing.

Lani built the SHOW Standard for indie, romance and romantasy authors, readers and platforms that want a shared vocabulary for what is actually inside a book.

Anshuman built it to stop coding agents from filling in the blanks and shipping a guess. If nothing is missing, it skips the questions and just does the work.

Nick recently quit his job at Google to work on Bisque full time, and asked for the audience segment pointing out that Andrew would be extremely early: the repo had one star, his own.

Kevin, whose magic-compact was on the Sep 11 show, built it for people who watch the channel, want AI to build something, then quit after installing six tools and hitting thirty error messages, especially on Windows. It is his most popular project.

Aron sent two tiny repos he thought fit the hidden-find radar. His point with MemHooks: long-term memory can store exactly the right fact and still fail, because the agent never thinks to look for it.

Yatin’s team wanted to know which photo to lead a listing with, rather than only A/B testing it, so he built the ranker, then open-sourced it because the approach is useful well beyond that one problem.

Alex built AgentScreen for people running Hermes across several computers, and says it lets you take control “just like Grokbot does.” He has no idea how to charge for it.

Bishal, a PhD who maintains ManuView, saw researchers paying consultants $39 to $1,800 and up per paper just to catch the flaws that get papers desk-rejected, while generic chatbots are too agreeable and make up citations.

Andreas, co-founder and CTO of SigniVoice, watches the weekly GitHub roundup and built GlanceFlow for the non-engineers on his team, who had started using Claude Code and found it overwhelming.

Christopher, author of the novel Dealing with Murder, heard Andrew say on the show that he was not sure how people give a whole book to an AI to follow, so he built Book Snapper as a proof of concept.

Christopher’s team at sthreelabs runs a small team of agents on Paperclip, from the Sep 25 episode, and kept wanting to answer agent requests without opening the dashboard, especially when he is remote and would otherwise set up a tunnel or wait until he is back at the office.

Fraol wrote in to say he had built a tool he wanted Andrew to look at, and offered to walk him through it.

James says his second repo has dropped, it is still in the early stages, and it is the first time he is adding some vibe coding into the mix, so “we will see how it goes.”

Ossie watches every Top 10 episode and built it on one rule: agents propose, typed code computes, so every number an agent quotes comes out of a tested Python calculator you can run yourself.

Donovan, a YouTube follower from Madagascar, based it on the Vakana beads practices of his Malagasy ancestors. He has just open-sourced it and would love the show to give it its first star.

Simon has been watching Andrew’s videos for ages and is building a whole host of completely offline Android apps. plain-vault is the first one he feels happy sharing, with a few more ready soon.

Lisa asked Andrew and Corey to review Switch, whether the agents can genuinely work together, how much setup it takes, and how permissions and costs are handled.

John, founder of BlackSnow Intelligence with a background in cybersecurity governance, risk and compliance, noticed that AI code-review tools and scanners can themselves leak source code, credentials or sensitive context to outside services.