Gregory F.

@gf

100 Following

41 Followers

Gregory F. pfp

The chatgpt api release make me consider saving $20/month just because 🙈 https://i.imgur.com/fUNQyXC.png

0 reply

0 recast

1 reaction

Gregory F. pfp

That is the wording that Meta uses when it wants you to enable 2fa 🙈🧟‍♀️ https://i.imgur.com/3ysiC7p.jpg

0 reply

0 recast

0 reaction

Gregory F. pfp

Balicaster https://i.imgur.com/Lhmogci.jpg

1 reply

0 recast

0 reaction

Gregory F. pfp

1-4? I will abid... https://i.imgur.com/GaMzbcA.png

4 replies

0 recast

0 reaction

Gregory F. pfp

What's still missing for a Chat-gpt-like-based AI freelancer working independently e.g. on upwork, that insist on text-based communications? When will it appear?

1 reply

0 recast

0 reaction

Gregory F. pfp

Tsugite is the art of Japanese carpentry joins, made without nails or glue. The precision required is staggering, and is mostly invisible from the outside. https://i.imgur.com/irgjlWM.gif

0 reply

0 recast

0 reaction

Gregory F. pfp

Does anybody know how much "context" can chat gpt hold in a "session"? It doesn't answer that question itself 😄

0 reply

0 recast

0 reaction

Gregory F. pfp

I'm now convinced ChatGPT can be used instead of a junior developer now. Input: a defined task + 5 lines of feedback. Output: a caching module in TS with a Redis backend, that support "stale while revalidate" etc. I then asked it to give feedback to itself, and it updated it based on that own feedback!

0 reply

0 recast

0 reaction

Gregory F. pfp

I definitely prefer a "series of sprints" approach to work. 🏃‍♂️ We were leisurely onboarding a large customer, when their existing online store suddenly broke down. Rolled sleeves, 24h of all-hands work and we relaunched it on our platform. Happy customer + 2 weeks worth of improvement to the product in two

1 reply

0 recast

0 reaction

Gregory F. pfp

Contrarian take: until we get to AGI, most value derived from AI/ML will come from businesses developing their own models, and not on large models used "as is". Reason is that dimensionality of the data manifold is too high for most problems, and a task-specific model reduces it. Large models will be used as inputs.

0 reply

0 recast

0 reaction

Gregory F. pfp

Non-crypto interests: 😊 Applied study of happiness ♾️ Beauty of maths 👥 How people work, 🧬 How biology works 🇮🇹 Romance languages 🌮 Elevated tacos 🍸 Pisco sour ⛵️Sailing around the Med 🧘‍♂️🏋️‍♂️🧗‍♂️ … ( Basically I’m not much into crypto 😂)

0 reply

0 recast

0 reaction

Gregory F. pfp

Love it when you're onboarding one of the first users, and they're like "No worries, I'll watch some course on how to use your product" 🤪

0 reply

0 recast

0 reaction