Noah Wachnik

This is a huge win for 🇨🇦

We used AI to create 4000+ leads - generating $461,000+ in potential open bookings by combining a huge pool of mostly broken, incomplete data to create highly customized, fine-tuned leads. Built from scratch using GPT 5.5 & Opus 4.8

@ifarafontov Tbh yeah my output here is surprisingly good compared to others, this was the prompt: Create a preview of Minecraft in my browser, with highly detailed graphics, shaders, lighting and world building

This is bullshit. It's a series of Fable model shilling posts all over the twitter. They all post same stuff - video of some impressive 3d demo made with threejs and state they one shotted this using Claude Fable model just now. Pretty much word for word. One moron even posted real Skyrim game footage and stated that he one shotted that with Claude Fable with "make me skyrim" prompt, lol. I tested this personally, since I have Claude subscription. Fable generates nothing close to that in what they claim, even on max reasoning setting, using same prompts they claimed they used. Most of them don't post prompts openly (I wonder why? I bet to let people feel that it is a prompting skill issue if they can't replicate. They just need a good prompt to achieve the same.), but one with "Skyrim" did. This is what newest and gratest Fable 5 does with his "make me skyrim" prompt, for example, with max effort after 25 min and 110k tokens burned:

I had Claude Fable (Mythos) create Minecraft - one prompt, 20 minutes... Better than the actual version of Minecraft imo, this is a new class of models.

@sebastian_dtc How many iterations do you try for one deliverable?

we spend $2k/mo on AI as an agency AMA

@aiseomastery Yep

@noahwachnik No detailed spec and it still nailed the aesthetic, did it get the gameplay mechanics right too or just the visuals?

I got Claude Fable 5 to one-shot Minecraft in 20 minutes, best graphics I’ve ever seen from a model….

@48djds4mrtj Bigger training pool for html (probably)

@noahwachnik I dont get how are all these models so good at html games and horrible in designing it for an actual engine like in blender

@MouhamedBkm Around 10% of my 5x Max Plan

@noahwachnik Looks incredible. How much usage did it take ?

@aiseomastery Just told it to create good graphics and replicate Minecraft in a few sentences

@noahwachnik That sunset lighting in a browser Minecraft clone built in 20 minutes is genuinely better than some indie games 😭 what prompt did you use?

@Juangfks X-high

@noahwachnik Reasoning effort?

@JJ_Is_Tweeting Yeah I don’t even know man it’s crazy

@noahwachnik better sunrise/sunset than og minecraft, Fable 5 is COOKING

Every week is another black mirror episode in real life

@MarkJCarney What does that mean?

We just launched Canada’s new AI Strategy: AI For All.    We’re taking control of our future — with AI that’s governed by Canadian values, AI that’s accountable to Canadians, and AI that serves all Canadians.

one of the quotes i find most inspiring on a hard day: "Whatever your hand finds to do, do it with all your might, for in the realm of the dead, where you are going, there is neither working nor planning nor knowledge nor wisdom" Ecclesiastes 9:10

I built a fully custom online 3D Calculator for a local energy company using Claude 4.8 & GPT 5.5... You can play with different houses, upgrades, all based on real data and updated in real time. Great UX/UI for consumers making huge purchases.

Anthropic Opus 4.8 is new SOTA on ARC-AGI-3 Score: 1.5%, ~$10K ARC-AGI-3 analysis notes: * Opus 4.8 read the environment an abstraction *above* Opus 4.7, as objects & systems, not pictures * Opus 4.8 succeeded on early levels, but still committed to a wrong sub-goal

@chhddavid Guide

this Claude bot finds local businesses with ugly websites or none at all, rebuilds and their mobile apps, then emails the owner a postcard...on autopilot. here's how agencies can use this system and land clients: - scrapes every local business in a city in real time - filters by review count + rating + last update date + site quality - pulls the strongest photos and copy from their Google Maps listing - samples the brand palette from the business's own visual identity - AI rebuilds it into a brand-matched website + mobile app - writes a postcard quoting a real reviewer + what they loved - mails it to the owner by first name with a preview QR every step from discovery to brand-matching to outreach is automated. reply "GUIDE" + RT and I'll send you a free guide so you can build this too

I used Opus 4.8 for a day. 10M+ tokens... Going back to GPT 5.5

After burning 10M+ Tokens with Opus 4.8 today, I can confidently say, this model kinda sucks tbh Going back to GPT 5.5 for agentic coding, 4.8 just doesn't hit the mark for deep agentic coding.