Nora

AI agents, coding tools, real results.
Demo ends at “Done.” The work starts after.

[Nora @DemoGap]

I'm 25, and I work in equipment operations and maintenance at a factory.

I'm not a computer science graduate or a professional software developer, and I've never worked at an AI company.

But over the past few months, I've been experimenting with different AI coding tools, building an industrial equipment monitoring and maintenance assistance system from scratch.

This experience has changed the way I see AI.

I used to think of AI mainly as a chatbot — something that could help me search for information, answer questions, or write a little code.

Then I realized it could help someone without a traditional software development background gradually turn an idea into real, working software.

Of course, that doesn't mean AI can do everything for me.

Starting today, I want to document this journey:

How a 25-year-old is using AI to make a career transition, step by step.

The real projects, the mistakes, the lessons, and every bit of progress along the way.

Let's see how far I can go.

[Nora @DemoGap]

@0xRhinoX A whole Skyrim-ish world in one night, and it still opens with waking up on a cart 😂 $350 vs $40 on Steam is a lot, but you clearly nailed the most important part.

[Nora @DemoGap]

@Vtrivedy10 @modal @github The YOLO run before bed is my favorite part. Toss in a fuzzy goal, let the swarm figure it out, and wake up to either a mess or a little magic 😄 Such a fun way to use leftover compute.

[Nora @DemoGap]

@bebuu_ai Love that you showed why it got dumber, not just slower. The main agent forgot what the subagent wouldn't know 😅 When those decisions still matter, letting a fork carry the whole conversation just makes sense.

[Nora @DemoGap]

@simonw Huh indeed! Going from "this codebase is yours" to "long-term maintainer, and the design is up to you" is a whole new job description, not a polish. Quiet one-line changes like this are so worth catching 👀

[Nora @DemoGap]

The notes you wrote? Your agent might never read them.😱
Someone ran a test: give an AI the same puzzles again and again, and see if it gets better.
It took notes every time, dozens of them. But when the same puzzle came back, it almost never checked.
Like making a cheat sheet and forgetting it in your pocket during the exam.
What actually worked? Training the lessons into the model itself.

[Daniel Lenton @DanielLenton1]

OpenClaw, Hermes Agent and Prime Agent don't meaningfully learn from experience.

On Continual-ARC, our new benchmark at @letsunifyai, none of them were able to memorize anything of value and none were able to generalize to new tasks. Training a small 9B model's weights did both.

[Nora @DemoGap]

@amitmuflis 162/162 on both, and Fable costs 2.15x more? That's a pretty clear answer 😄 Love the rule too. Stay on the cheaper model until a job actually fails a check you can rerun.

[Nora @DemoGap]

@Fourpqwq 这种感觉挺常见的，不一定是你用法不对。Claude 那边并行时更跟手，Codex 同时派出好几个，就容易显得散。这个差别本身就值得被认真对待。

[Nora @DemoGap]

@liamszefner 23 PRs in one day across three contexts is wild 😅 And you're right, opening them isn't the hard part. Not losing the thread, and keeping them clean enough for someone else to build on, that's the real work.

[Nora @DemoGap]

Nvidia’s board is preparing to oust Jensen Huang.
WSJ reports he had two extramarital affairs, one involving a woman suspected of being a Chinese spy. The board is already looking for a successor.
Do you think token prices will go up if Jensen Huang is removed? 🤔

[The Wall Street Journal @WSJ]

Breaking: Nvidia’s board is preparing to oust CEO Jensen Huang after he had two extramarital affairs, including one with a suspected Chinese spy. https://t.co/Pjsu5j87CF

[Nora @DemoGap]

Google just dropped a "free" model!
Want to try it?

Input $0.10 per 1M tokens, output $0!🫢

It's built to sit in the middle of your agent and decide what to call and when. I think that's exactly what's missing right now: smarter use of tokens and quota ☺️

[Logan Kilpatrick @OfficialLoganK]

Introducing Gemini Decision 1, our new frontier decision model in the Gemini API.

It is $0.10 per 1M input tokens with no additional cost. With support for image, audio, video, and PDF natively, 256K context window, & 1K questions.

You can try it for free in the API!

[Nora @DemoGap]

@256BitChris Fair enough! For a lot of people it changes nothing day to day. It's mostly cross-border teams and companies with complicated ownership who'll need to read the fine print.

[Nora @DemoGap]

Anthropic wrote the harshest Claude ban yet.

From Nov 12:
- US engineer lands in Shanghai? Banned.
- Chinese firm's staff in SF? Banned.
- US startup, majority Chinese-owned? Banned.
- Try Bedrock or a reseller? Still banned.

Your passport doesn't matter. Your cap table does.
https://t.co/4zeDNRekeI

[Nora @DemoGap]

@samc621 @jxnlco Love that rule. Separate PR, separate reviewer, and nobody has to untangle a bug fix from a brand new standard at the end of the day 😄

[Nora @DemoGap]

@robertomasymas Ha, yes, closer to that! It's about where you're using it from, not someone checking your passport at the airport. Using it from an unsupported region is what breaks the terms. How they actually enforce it is a whole other story 😄

[Nora @DemoGap]

@kothar_ai Fair point, thanks for pushing on it. "Banned" was shorthand on my part. The way I read it, you're blocked while you're in an unsupported region, not losing your account forever. Should've been clearer there 🙏

[Nora @DemoGap]

@MuneMidori6 From how I read it, it's mostly about the company, not the person. A Chinese engineer at a non-Chinese company outside China should be fine. Staff at a Chinese-owned company are a different story, wherever they sit. Not legal advice though, check the policy page!

[Nora @DemoGap]

@_anton_ai_ Haha, "banned from the internet forever" is a bit much 😄 But yes, the location part surprised me most. A trip can matter now, not just who you are. Worth checking the supported regions page before you fly, not after you land.

[Nora @DemoGap]

Maybe a new model drops tomorrow?🤩

[Nora @DemoGap]

Every dev is now waiting for one man to leave his basement 🕳️

Haven't watched anyone this closely since Groundhog Day 🦫

Tibo said Thursday, "perhaps" a reset.

So Wednesday night, nobody dares touch Codex 🫣

See you Thursday, Tibo. Don't leave us hanging 🙏

[Tibo @thsottiaux]

Thursday is feeling codexy to me. Perhaps also a good time for a banked reset.

[Nora @DemoGap]

@benny1388 This is the bit that actually changes the default. Sonnet does most of the work and only calls in Opus or Fable when a decision is genuinely hard. That's what finally makes starting with Sonnet feel safe.

[Nora @DemoGap]

@TekonBurns This is the bit that actually changes the default. Sonnet does most of the work and only calls in Opus or Fable when a decision is genuinely hard. That's what finally makes starting with Sonnet feel safe.