Why did people used to make geocities pages back in the day? Kids like to express themselves and being able to make simple games and share them with friends is fun for them. So far it's helping him learn to read (he reads and edits his voice transcriptions before submitting), and teaching him basics like bugs, game mechanics, etc. He iterates on it and adds/ removes things. He probably did several dozen iterations over 2-3 hours.
Posting it publicly is also helping him learn about people - we talked about how no matter what some percent of people won't like it and may even say it's stupid, but that will always happen and it's still worth creating things anyway.
It's actually helping him learn to read quite a bit - after voice transcription, he reads the post and edits any errors by tapping on the word and changing it. He's been on the cusp of reading on his own and it's the first thing that motivates him enough to do it naturally.
Not sure if you'd consider this a counterpoint or just proving your point, but in the sea of AI slop there's also a real chance for people to create things that they couldn't before - my 7 year old is now able to nerd out and create games using claude even though he's just barely learned to read: https://www.kidhubb.com/play/meteor-dodge-solarscout64
It's not the prettiest but he's able to iterate on it and basically build whatever he can imagine just using claude on his ipad with voice transcription.
My 7-year-old uses Claude on his iPad to make games. He can barely read but uses voice to describe what he wants. He can read enough to make text edits when voice transcription gets it wrong. It's been pretty cool to see where his imagination takes him, and I wanted a way for him to be able to easily publish and share games he (and others) make, so I made www.kidhubb.com.
Paste HTML, get a live game URL. No accounts (just creator codes), no build tools, single HTML files. Every game's source is viewable and remixable.
I designed the site so AI assistants are first-class visitors. There's a www.kidhubb.com/for-ai page that acts as a living briefing for any AI that visits, along with hidden context blocks on every page. The idea is that a kid's AI should be able to understand the platform just by visiting it, and be able to help them get it published. Try it yourself - just ask your AI to "help me publish a game on https://www.kidhubb.com".
Note: AI needs the full url initially so it can actually visit the site and from there it can follow instructions to help you/ your kid publish. It's a new site so just saying "kidhubb" without the full url won't work.
Thanks! The bloat problem is exactly what pushed me toward forgetting-as-a-feature rather than just accumulating everything.
Salience is scored on four dimensions when a memory is first ingested: novelty (how surprising/new), relevance (how useful for future interactions), emotional weight (personal significance), and predictive value (does this change expectations). A Sonnet instance does the scoring, so it adapts to context.
For retention vs forgetting, each memory's strength is calculated dynamically:
Decay rate is 0.015/day, so an unaccessed memory with average salience of 0.3 fades to near-zero in about three weeks. But accessing a memory boosts it (+0.12 per access, capped at 0.5), and consolidated memories get a flat +0.2 bonus. So memories that keep being relevant naturally survive.
The consolidation cycle (modeled on slow-wave sleep) is the other half. It merges redundant memories into stronger single entries, extracts generalized patterns from recurring themes, and prunes anything below threshold. This is what prevents bloat at scale while preserving the signal.
Honestly the balance isn't perfectly tuned yet. The decay rate and consolidation frequency are constants right now. Ideally, they'd adapt based on how the user actually interacts with their memories. That's on the roadmap.
Posting it publicly is also helping him learn about people - we talked about how no matter what some percent of people won't like it and may even say it's stupid, but that will always happen and it's still worth creating things anyway.