Sander Jansma

Optimistic maker of useful & playful products. Apps, tools, games, prototypes, often with AI. Building Dad², nerd-culture, gummybear.

[Sander Jansma @sanderjansma]

The world is better when people make things.

Useful tools. Playful apps. Games. Strange prototypes. Happy little inventions.

AI helps more ideas become real enough to try.

More creativity in motion. More courage to start. More joy in the world ❤️

[Sander Jansma @sanderjansma]

Why Terry? Because, Terry Pratchett is amazing!

[Sander Jansma @sanderjansma]

Yes, AGENTS.md exists. But in long agent sessions it doesn’t always stay in the live reasoning loop.

So I write “Terry” in my prompt.

The agent knows that means: apply my build standard: no ceremony, clear ownership, agent-operable docs/logs/CLI, E2E proof, no dead code, etc.

[Sander Jansma @sanderjansma]

I have the feeling OpenAI is silently and sneakily updating the GPT-5.5 model to become better than Fable, but we won't really know for sure since it's done in the background. No grand release means no fuzz!

[Sander Jansma @sanderjansma]

I just don't have enough tokens to build my autonomous research and writing app, my OS Rust ai-first game-framework and engine, my AI social sentiment benchmark, my ai-first todo-and-note-taking app, my "ai layer that helps people shape life around intent, context, and care", ..

[Sander Jansma @sanderjansma]

I can't wait for the day that OpenAI models (and Codex) get good at design. So, today? :D

[Sander Jansma @sanderjansma]

Now that loops are a thing, should we now move more towards a system where a plan is build multiple times in parallel and the best parts of each build are brought together in one final build - reaching the intent as close as possible? All autonomously of course.

[Sander Jansma @sanderjansma]

@wesbos Stroopwafels?!

[Sander Jansma @sanderjansma]

@AnthropicAI It’s bizarre that this moment already happened so fast in time. Not from a governance perspective (I understand the decision), but from a technical perspective. Really gives a glimpse of what’s in store for the future

[Sander Jansma @sanderjansma]

So, what are everyone's thoughts on Claude Fable 5? I've just added it to https://t.co/iV00DcuAOr, and can't wait to view all your sentiments on it! Care to help me out with this? :D

[Sander Jansma @sanderjansma]

Congrats @AnthropicAI on Claude Fable 5.

Now I’m mostly just sitting here wondering what @OpenAI does next.

What a time to be AI-live.

[Sander Jansma @sanderjansma]

I feel like the best addition to Codex is better management of sub-agents, agent teams and sub-sub-agents, or a better connectedness of conversations and cross-conversations. This is where Anthropic/Claude Code still beats OpenAI/GPT/Codex imho

[Sander Jansma @sanderjansma]

@bridgemindai Jep, agreed! This is why I’ve build https://t.co/iV00DcuAOr, since benchmarks only show but a section. It is human sentiment what’s really interesting: why use which model, in which harness, for what job? And why switch or stay?

[Sander Jansma @sanderjansma]

@thsottiaux Oeh, I am building https://t.co/iV00DcuAOr, how AI is doing today according to the people that use it; https://t.co/BBkIhco91B, an open-source note-taking app with todo's and AI-orchestration, a mobile game that will be a mixture of Diablo/Brawlstars/Hades! And more :o! Pick me!

[Sander Jansma @sanderjansma]

Could anyone explain to me why I should be writing loops and not prompts by giving a great example (except the “babysit PR” one)? I’m looking for the most interesting use-cases!

[Sander Jansma @sanderjansma]

@imjustnewatai Is this manifesting?

[Sander Jansma @sanderjansma]

@TTrimoreau The people who survive are the ones who can decide what’s worth making.

Code gets cheaper. Taste, judgment and useful personal tools get more important.

[Sander Jansma @sanderjansma]

Wow, Guild Wars 3! Can’t wait 😱

[Sander Jansma @sanderjansma]

@ThePrimeagen AGI!

[Sander Jansma @sanderjansma]

First, we wrote code for others to use.
Then, we built tools for agents to use.
Then, agents started writing our code.

Maybe next is personal tooling: people shaping their own layer of tools, agents and processes around how they think, work and live.

[Sander Jansma @sanderjansma]

Assume AI has “solved” coding. What’s the next paradigm?

[Sander Jansma @sanderjansma]

I feel like we will move more towards intent based interaction, where you shape and configure the world around you to enable outcomes and goals. Not via “apps” created by someone else, but with personal crafted systems and interfaces.