Login
You're viewing the assemblag.es public feed.
  • Marco Armentmarcoarment@mastodon.social
    Aug 28, 2026, 7:54 PM

    RE: social.tupo.space/@joshua/1171

    My current AI-code policy is that I will always use the best model I can access, as long as it's practically able to do the job.

    Right now, as far as I know, that's Fable.

    Quality and accuracy matter FAR more to me than volume or speed of output. And I feel that I owe it to my customers — and myself — to use the best tools at my disposal.

    Cost may be a factor to others, so no judgment. This is just my policy for myself.

    💬 5🔄 0⭐ 0

Replies

  • 💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Marco Armentmarcoarment@mastodon.social
    Aug 28, 2026, 8:49 PM

    @callin @dimitribouniol Yes, the cost is pretty different. I can't hire a trustworthy programmer for $100/month.

    Fable also isn't LIKE another human programmer. It's great at things I'm bad at, and bad at things I'm great at.

    Another human programer would have a lot more overlap with my skillset, which I really don't need.

    💬 3🔄 0⭐ 0
  • Dimitri Bounioldimitribouniol@mastodon.social
    Aug 28, 2026, 9:09 PM

    @marcoarment @callin As someone who is seeing nothing but the harms this technology is bringing (especially for the next generation), it saddens me to read that you see things this way.

    💬 1🔄 0⭐ 0
  • Jon Mullinsjxmullins@hachyderm.io
    Aug 28, 2026, 10:11 PM

    @dimitribouniol @marcoarment @callin It worries me more there’s an argument for Marco to spend $100 to hire an imaginary person to do the jobs he wouldn’t do otherwise or be freed up to do.

    Especially with the St Jude’s campaign coming up where $100 can make a huge difference. Matter of fact we need the Marco Offset as well as a Max AI offset where people donate the same amount that they paid for AI costs this month.

    💬 0🔄 0⭐ 1
  • Dimitri Bounioldimitribouniol@mastodon.social
    Sep 5, 2026, 6:06 PM

    @marcoarment @jxmullins @atpfm Just wanted to write back in to say I’m happy to hear of the new proposed Yearly AI Marco Offset tier ❤️

    No clue how much folks actually spend on such things because I use none of it, but here is my extrapolated $200/mo → $2400/yr offset to offset the offsets even further

    A $2400 donation to St. Jude
    💬 1🔄 1⭐ 1
  • 💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Jake 👨🏻‍💻codejake@mastodon.social
    Aug 29, 2026, 3:52 PM

    @marcoarment @callin @dimitribouniol Great points, Marco.

    The other point I'd make is that Fable and agents generate (usually) better code *much* faster than humans.

    In my scope, this means I can spend more of my time with people. Any reasonable person has reservations about AI.

    But, isn't this what we use computers for? To do things humans can do, but faster and better?

    💬 0🔄 0⭐ 0
  • Óscar Morales VivóMyLittleMetroid@sfba.social
    Sep 5, 2026, 6:52 PM

    @marcoarment @callin @dimitribouniol We have the metered enterprise plan at work, so we know exactly how much each of us spends on the 🤖.

    I have teammates who legitimately spend more than a great engineer in Eastern Europe would cost or an ok one in Western Europe.

    We're in Silicon Valley so it still pays off for us but I don't see how this scales considering our bills aren't going to cover Anthropic costs either way.

    💬 0🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Wadewadiest@mas.to
    Aug 28, 2026, 9:58 PM

    @marcoarment Note that Opus 5 outperforms Fable 5 in some tasks, according to some benchmarks and comparisons. Fable 5 is better overall, but (within the dev domain) does seem tuned towards bug finding in particular.

    But IMO the big reason to not always use Fable (or Opus, over Sonnet) is both cost and speed. Asking Fable to do straightforward tasks is tedious; Sonnet 5 can do a lot of things just as well and in a tenth of the time. And iteration speed is important to overall productivity.

    💬 0🔄 0⭐ 0
  • Hendrikhendrik_kueck@mastodon.social
    Aug 28, 2026, 11:06 PM

    @marcoarment Give GPT 5.6 Sol a try and compare (maybe give both the same task). It seems that Fable is smarter in general, but specifically for iOS/macOS code, Sol is significantly better. The 20$ plan is enough for test driving it.

    💬 0🔄 0⭐ 0
  • neverbelievedep@mastodon.social
    Aug 31, 2026, 4:47 AM

    @marcoarment I don’t know what you are up to these days. But if you are not trying to mitigate AI externalities by setting up your own local code assistant using a beefed-up Mac Studio.
    Who will ?
    It is a fascinating tool. I don’t see how it could go away. But it does not mean we should surrender to the way it is currently marketed.

    💬 0🔄 0⭐ 0