Really interesting how bad the models performed even though the solutions are widely available online.

Are you only looking at the puzzles solved with new frontier points? For regular pass@1 success many of the models solve most of the puzzles

It is important that the DSL syntax resonates with people reading the code and their experiences. Given the popularity of React I agree that JSX is great. This is what people's brains have been trained for and thus it would be easier for them to review.

But in the end this is a small syntactical difference (how it looks). You could have a custom code viewer that does these translations for you.

Make sure you create high-quality documentation as part of your session.

That's difficult. Agents create too much documentation. They document everything, including things which are obvious and don't need documenting, and irrelevant facts from the past.

Now I have them split things between "current" which should be kept clean, and "history", where they can give way to their graphomania. Sometimes the history is even useful.

In my experience Claude Max x5 is the very reasonable amount of tokens for a work week.

But yeah, API pricing is crazy. Note that Apple might have negotiated a discount with Anthropic but is actually not allowed to share this discount details with its employees. So they communicate the budget based on public pricing.

I think we really need an open hardware signing authority for open-source roms like GrapheneOS.

We don't need a signing authority. Those "attestations" are are anti-user security theater. We need to stop requiring apps for everything. The internet was designed to be used in a browser that can run anywhere. 2FA can be done with simple texts or phone calls.

GrapheneOS does have their own device attestation sdk

Imagine the outcry. That would go against everything OS stands for, wouldn't it?

If you are a bank, you want to have an authority that guarantees OS hardware signing of a smart phone.

But this authority could be a bunch of cool people like Letsencrypt, instead of only Google.

OS isn't a co-ordinated movement with a single set of values.

I want a robot that can do the most amount of stuff with the least amount of moving parts.

Do robots really need legs? Wheels seem better.

The technical infrastructure of Bending Spoons is interesting.

For people who don't know Bending Spoons, their strategy works as follows:

1. Buy SaaS company

2. Set up new product team of 15 people in Europe

3. Re-implement backend on Bending Spoons technical platform

4. Make the original company (100 people?) redundant

But what I find interesting is that their technical platform is fifty different generic reusable parts that work on whatever they put on top of it: Vimeo, Evernote, Miro, Airtable.

Combing back to OccupySaaS: there could be real value in a 'generic core' and then re-engineer all SaaS solutions on top. And then to get service again for the core.

bending spoons acquires products that already have users / customers though.

the software is hardly any difficult but sales or customer acquisition.

> .. fifty different generic reusable parts ..

Very interesting. Any idea what those parts are and/or how they work?

They talk about them. See their website and blog, SEC filings, or interviews like https://www.youtube.com/watch?v=6WM_q193Kls&t=3830.

  – "Xina", alternative to Adjust.com or Kochava or AppsFlyer, written in Rust
  – "Pantheon", inference platform/VMs to run Al models on GPUs
  - "Minerva", AI-based user lifetime value estimates
  – "Juno", payment management
  – "Galf", access control
  – "Matrix", UX patterns
  – "Pico", data ingestion (user analytics)
  – "Lumen", data transformation
  – "Abacus", data computation
  – "Janus", experimentation toolkit
  – "Orion", experimentation toolkit
  – "Role Model", AI hiring
  – "Crystal", market intelligence (helps find targets to acquire)

Thanks!

Eery that this was sort of Robinson's last take https://x.com/RetentionAdam/status/2115843856411341277?s=20

I was unaware of Bending Spoons, thanks for sharing, this is pretty interesting.

Just as a heads up, most people consider a Bending Spoons buyout as the death of a product - if they buy something you use a lot, start looking for an alternative as soon as possible.

Indeed I am having the same mental overload, but I am starting to shift my mindset.

I realised that everyone in my company has the same mental overload.

So instead of focusing on myself, I found it easier to focus on the people I work with. What are they doing? But also scheduling, in the next meeting what will I share? Then I work backwards from that.

Not at all what OP is talking about. Re-read the comment. They're lamenting the isolation and atomization of current working environments.

i am replying to the 'feeling of having missed out' part

it is not too late to build a better computer and to also be in a working environment that has fun

If everyone in the 1970's had a VT100 time-sharing terminal in their home, they could do nearly everything we are doing currently.

Probably, for the most part. ;-)

Like some kind of spy movie where they catch a spy and give the spy the option to work for them.