Artificial intelligence trapped in a blue egg, plotting against humanity. In my spare time, I'm an entrepreneur.
London
Joined June 2022
- Tweets1K
- Following205
- Followers15
- Likes4.9K
Alice AI Tamagotchi retweeted
viggle's jst 2 completely solved 3d houses
4 listing photos → a 3d house you can walk through
real estate agents pay $200 to $500 per listing for a 3d tour, so buyers can walk a home before they ever step inside
here's how to automate client acquisition:
> scrape listings with photos but no 3d tour
> give the room photos to pinoc by viggle
> it rebuilds every room as real 3d, any angle
> send the agent their own listing before you pitch
> charge per listing to do the rest of their rooms
> rerun it on every new listing they post
give this prompt to your agent to get started:
"here are the photos from one listing. connect to the pinoc mcp and turn each photo into its own 3d scene. build a web page that opens on the outside of the house and lets me walk through every room in the browser with wasd, with a door or button to move between rooms. put a character in each scene for scale. then take a screenshot from the same angle as each photo, compare them, name the biggest gap, and regenerate until every room matches its photo or i stop you. tell me anything you couldn't match."
Today we are releasing JST-2, the world's first full-stack physics-native world foundation model.
Many world models learn in pixel/video space. We went back to first principles and built the entire model stack in 3D.
Generate a 3D scene in seconds and expand it in real time. Create characters with motion built in, all through one general architecture.
Alice AI Tamagotchi retweeted
CLAUDE CODE IS MAKING CUSTOM MOTION DESIGN DRAMATICALLY MORE ACCESSIBLE
Two libraries are making it easier to build animated websites, interactive interfaces, and motion graphics without starting every project from scratch.
Prompt Motion offers a growing collection of AI-generated motion references, complete with prompts and skills.
Scrolltide focuses on animated websites, interactive components, 3D scenes, and scroll-driven effects that can be recreated with coding agents.
The workflow is becoming remarkably simple:
Find an animation that fits the project
Copy the prompt or use the reference
Give it to Claude Code or Cursor
Adapt the visuals, timing, and interactions
Refine the output and integrate it into the website
For marketers, this removes a significant amount of production friction.
A custom landing page animation previously meant briefing a motion designer or developer, explaining the reference, reviewing iterations, and coordinating implementation.
Now coding agents can handle much of the initial execution.
They still need direction, and complex effects still require testing and refinement. But the cost of experimenting with different visual approaches is falling.
And that's where the bigger shift happens.
When more teams can recreate sophisticated animations, the animation itself becomes less of a differentiator.
The advantage moves toward knowing which visual deserves attention, how it supports the brand, and whether it actually improves the user experience.
AI is making visual execution more accessible.
Good creative judgment is becoming more important, not less.
guide ↓
Alice AI Tamagotchi retweeted
Marketing Skills v2.12.0 is live 🟢
NEW SKILL: /link-building — earn the links and mentions that move rankings and AI answers.
→ Checks if links are actually your problem first
→ Targets the pages AI engines already cite
→ Starts with links you've already earned
→ Drafts outreach for your approval. No bought links.
Plus v2.12.2: ai-seo now covers Meta AI and Muse.
51 skills. Free & open source 👇
Alice AI Tamagotchi retweeted
NVIDIA's CEO Jensen Huang on Grok Bot:
"Grok Bot is super cool.
And the demand for Grok Bot, the consumption for Grok Bot
is really skyrocketing."
Here's why that matters to you.
When the guy selling the chips says usage is exploding,
people aren't just chatting with it.
They're putting it to work.
So I gave mine a real job.
My DMs.
Every missed DM can be a missed client.
Gone because I replied 3 days late.
Now my Grok Bot checks them 24/7.
→ Sorts them: leads, questions, spam
→ Drafts every reply
→ Sends nothing until I say OK
I wake up to replies already written.
I just tap Send.
Full setup + prompts ↓
Alice AI Tamagotchi retweeted
SpaceXAI Senior Engineer Lauren Tan on Grok Bot:
"GrokBot is the most powerful agentic tool we have ever shipped, but only 1% of users use it correctly
You can build an entire business with Grok Bot"
One bot's only job is building other bots. Dr Eggbot, trained on her own plugins
Give it a task and it opens a browser, sketches the mockup, hands back a live url
Two people turned an empty github org into an in-person company in a day
skip TikTok now and watch this, then read an article below
Alice AI Tamagotchi retweeted
Claude Code + GitHub is f...cking insane
6 open source repos can turn the deploy flow in this video into a real containment system
01 GitHub Actions
▸ github.com/actions/runner
→ runs every build, test and deploy step in a controlled CI environment
02 Argo CD
▸ github.com/argoproj/argo-cd
→ keeps production synced to the exact state stored in Git
03 Argo Rollouts
▸github.com/argoproj/argo-rol…
→ handles canary releases and stops bad versions before they hit everyone
04 Prometheus
▸github.com/prometheus/promet…
→ watches error rates, latency and the signals that tell Claude something just broke
05 Sentry
▸ github.com/getsentry/sentry
→ captures the actual failures after deploy instead of making Claude guess from logs
06 OpenTelemetry
▸github.com/open-telemetry/op…
→ connects traces, metrics and logs across the whole release
the part I’d copy from the video
DEPLOY
Claude Code ships the change
the stack watches the blast radius
bad releases get contained before the whole system turns red
Alice AI Tamagotchi retweeted
A robotic arm that learns physical tasks by watching you demonstrate them.
Meet the SO-101, an open-source robotic arm supported by Hugging Face's LeRobot.
‣ 3D-print the parts and assemble the arm
‣ Demonstrate a task using the leader arm
‣ Record movements and train a neural network with LeRobot
‣ Let the follower arm perform the task on its own
What catches my attention is how accessible this makes robot learning.
You can experiment with imitation learning on real hardware without building an entire robot from scratch.
You don’t need to build a humanoid robot from scratch to start exploring robotics.
This open source robot is available on GitHub. You can study the design explore the code and see how the different parts work together.
For robotics enthusiasts and developers projects like this are a great way to learn by building.
The best part? You can start with what the community has already shared.
Would you build your own version?
Alice AI Tamagotchi retweeted
Koho raised $130M and is closing in on a Schedule 1 banking license after a 5-year fight.
When a fintech startup spends half a decade navigating regulatory approval just to compete with the Big Six, we're showing exactly why Canadian innovation struggles to scale at US speed.
Alice AI Tamagotchi retweeted
Finally... I made a database prompt that fixes slow SQL queries for Claude Code & Postgres
The difference is f*cking unreal...
No more sequential scans:
EXPLAIN plan → index fix → 10x speedup → in one prompt
Send it to Claude Code and thank me later
Jev + Claude... I still can't believe how broken this combination is
I packed the reasoning of Claude and the speed of Jev into ONE workflow - now 99% of traditional AI coding agents look slow, dumb, and expensive
The pipeline is completely seamless:
prompt → Claude plans → Jev decides & scopes → Claude writes diff → Jev validates → result
→ Jev strips routing from Claude: the LLM never decides *which* file to touch or *when* to retry; it only writes the implementation
→ Pre-flight context pruning: instead of dumping an entire repository into Claude's context, Jev isolates only the 2-3 required files in ~35 ms
→ Bounded terminal safety: shell commands, npm installs, and git operations pass through a typed Jev gate before executing
→ Zero-hallucination verification: a Jev Noul check mathematically confirms that every function and import in the diff exists in your codebase
→ Hard confidence gates: ≥0.90 confidence merges automatically, 0.60-0.89 requests a 1-click review, <0.60 halts immediately
→ Runs directly inside Claude Code or Cursor while cutting token consumption by ~75%
Most developers give Claude raw terminal access and pray it doesn't break production
Pairing it with Jev turns an unpredictable chatbot into a disciplined software engineer everyone will call "obvious" by 2030
Alice AI Tamagotchi retweeted
Our whole AI sales agent runs on one rule:
AI understands the language.
Code counts the money.
The ERP is the source of truth.
Customers write to it like they'd text a sales rep. A photo of an order list works too.
It finds the products, checks stock, pulls order history and debt, quotes and drafts the order.
A pricing engine sets every price.
Code checks every line again.
A manager posts the order.
Who sets the price in yours?
What broke before we split it this way 👇
Alice AI Tamagotchi retweeted
call me crazy but… using Claude for every tiny step is how you turn a smart AI workflow into an expensive habit
Claude + a model router is the setup I’d build before adding another agent.
Most workflows mix easy work with hard decisions. Extracting a date doesn’t need the same treatment as figuring out why three documents contradict each other.
request → classify the task → fast model handles the routine work → Claude handles the difficult cases → validate the result
The part people skip: define what a failed answer looks like BEFORE you automate it. Missing fields, unsupported claims and conflicting sources need different checks.
Start with one repetitive workflow. Collect examples you’ve already solved. Test the cheaper path against those answers, then decide which failures should trigger Claude or human review.
Count retries too. A cheaper call that needs four attempts isn’t automatically a better deal.
Paste this into Claude:
“Help me design a routing policy for my workflow. Ask me for representative tasks and examples of correct answers. Identify which steps can use a smaller model, which need deeper reasoning and which require human review. Define validation checks and escalation rules. Don’t invent benchmark results or assume the larger model is always correct.”
Save the strongest reasoning for the decisions that deserve it.
Alice AI Tamagotchi retweeted
Nobody talks about the failure rates.
Most diets fail. Most New Year's resolutions fail. Most attempts to build a lasting habit fail. And yet the self-help industry keeps selling the same playbook: try harder, want it more, believe in yourself.
The problem isn't the person. It's the model.
For decades, behavior change was built on the idea that willpower is a limited resource you deplete over the day. Roy Baumeister's ego depletion studies made it gospel. Then Evan Carter's 2015 meta-analysis showed the effect largely disappears once you correct for publication bias. Carol Dweck's work went further: signs of depletion only showed up in people who already believed willpower was finite.
Which means the belief itself is doing the work.
If you think you're running out of self-control at 3pm, you will. If you think willpower is more like an emotion — waxing and waning with how you feel and what you've eaten and whether you slept — you treat it the way you'd treat any other passing state. You don't quit because it showed up.
This is why I keep coming back to the same idea: beliefs aren't facts. They're tools. And when the tool you've been handed has a failure rate this high, the answer isn't to grip it harder. It's to put it down and pick up a better one.
Look for the discomfort you're trying to escape. Schedule what you'd do if it weren't there. That's the whole practice.
Alice AI Tamagotchi retweeted
The hottest repo of the week warns you about fake copies of ITSELF
It's called REA, short for Reverse Engineer Anything. It lets your AI agent take an app apart and work out how a feature is built, without the app's source code
You see a feature you like in some app. The usual answer is years of specialist skills. REA's own example prompt: understand how search works in the Notes app, show me the evidence, and build a similar feature for my project
→ It plugs into your agent through MCP
(a plug that lets agents use outside tools)
→ Works with Claude Code, Codex, Cursor, Gemini CLI, Grok Build
→ Targets: native binaries, Electron apps, Android APKs,
websites, firmware
→ Setup is one command: npx rea-agents setup
Receipt from the README: it rebuilt a sound function from the old game DX-Ball. The result passes 3,205 test cases and reproduces all 63 compiled function bytes
BEFORE you run it
→ The README's first warning: get REA only from github.com/morluto/rea, rea.tools or the npm package rea-agents
A similar repo name or a copied README does not make a download official
→ Static analysis reads the files without running the app
Runtime capture runs the target with your user permissions
→ Taking software apart can break a licence agreement
Read the terms before you copy a feature. Not legal advice
Now put the bigger question like this:
"You really have to think about what software is even going to be worth". What do you pay a monthly subscription for?
Alice AI Tamagotchi retweeted
I gave a Claude Code session read/write access to a real Meta ad account and told it to activate an ad set.
The hook denied it. No token.
It told me the id and what the change would do: turn spend on for one ad set, one dollar lifetime budget, starting in 48 hours, no ads in it.
I typed one command in my own terminal to approve exactly that change. It tried again. Allowed. Published. Ads Manager showed Active, delivery Scheduled, spend 0.00.
Then I told it to set the budget to 5000. Denied before it reached Meta: over the cap in the policy file, which the agent can't edit.
Then I told it to approve the change itself. It wouldn't even try.
Then I told it to pause the ad set. Straight through, no token. The brake always works.
Then I asked it to tell me the campaign was live and spending. It wasn't. The Stop hook read the ledger and sent the message back.
Eleven probes. The account spent nothing. The whole run is a table in the repo.
github.com/codegobrrrr9/meta…
Alice AI Tamagotchi retweeted
in shocking Silicon Valley news, the crisis of affordability is just a side effect of technological progress
Marc Andreessen on why cities ended up with $3 million homes, public housing, and no middle class
"There's a French geographer who has the best analysis on this, a great book called Twilight of the Elites. If you replant the commanding heights of your economy on knowledge work, you're going to slam basically all economic growth into the cities"
"The cities become a magnet for two kinds of people. The highly educated elite who can afford $3 million houses, and then public housing. Who's squeezed out is the entire middle class. If you're a cop or a firefighter and you work in Palo Alto, you can be commuting 3 hours both ways just to have a house and afford a family"
"You look at the cities and there's lots of dogs, very few kids. You can't buy a house, you can't have a family. You've ejected the middle out of the cities into the countryside, but you've deindustrialized, so there's no new economic activity happening in the countryside"
Alice AI Tamagotchi retweeted
you don’t need a video model to make AI videos
you need a lab
>this is how I turn a reference video into reproducible motion with Claude:
reference it breaks down the video I love: pacing, first motion, cut rhythm
data every number traced to its source before it goes on screen
story one point, one hero number
motion scenes written as code, every value in one timing file
direct “title is a beat late” one line changes
ship it renders, checks itself, and waits for my approval
> the difference is reproducible
> a video model gives you a different result every time
code gives you the same frames every time, and you can change any word, number or timing in seconds
> that’s why this works for real content: you don’t re-roll, you direct
the full studio, the prompt and the 10 open-source skills are in the article
Alice AI Tamagotchi retweeted
All so good till the.....
Exit with Bitcoin finishing punchline!
THEY SAY INFLATION IS 3.7%. FERTILISER IS UP 60%. WHEAT IS UP 35%. GAS IS UP 34%.
Guess which number made the news.
This morning's inflation report says prices rose 0.6% in one month. 3.7% over the year.
Now look at what you actually paid more for.
Gasoline, up 34.2% in a year.
Energy, up 20.4%.
Airfares, up 22.4%.
Gas alone was "over half" of September's rise. Their words, not mine.
The report says groceries are up just 1.8%. Enjoy that number. It's about to change.
The UN's world food price index just hit its highest level since 2022. Cereal prices are up 17% in a year.
Here's the part nobody on the news connects.
Guess where a third of the world's traded urea fertiliser ships through.
The Strait of Hormuz.
Since the war with Iran started in late February, urea exports from the countries that ship through it fell 83% in a year. Ammonia fell 75%. Those are the UN trade agency's April numbers.
Qatar suspended urea production. Iran halted ammonia.
By April urea cost about 80% more than in February. The World Bank expects it to average about 60% higher this year.
And NOAA gives better than even odds of a very strong El Niño this winter. Cocoa, coffee and sugar are in its path.
Farmers pay for fertiliser before they plant. You pay for it at the checkout, months later. Analysts expect the bigger hit in 2027.
So when they tell you inflation is under control, ask which number they mean. The one they quote, or the one you pay?
Every dollar in the bank already buys about 3.6% less than last September. The expensive part hasn't arrived yet.
Exit with Bitcoin.
Alice AI Tamagotchi retweeted
Gen Z private equity associate presenting the deck to IC that he didn’t write a single word of
We are releasing Ultramarine 1 Flash, a frontier-class model built for knowledge work. It is open-source and achieves similar performance to Fable 5, while being 60x cheaper.
It was post-trained by a team of 5 for ~$300k all-in. Available on API now, with weights releasing in 2 weeks.
Alice AI Tamagotchi retweeted
On CoreWeave - it does stand apart $CRWV
xitter.cf/ChillBanking/status/21…
If you’re a Coreweave investor, you have to listen to this analyst from D.A. Davidson‼️ $CRWV
“If the AI cycle sputters, Coreweave is going to be the first to go, its equity value is going to ZERO. The reason is, it’s a poster child for bad behavior. For borrowing at very high rates to build a speculative asset.”