Jack Roberts

🤖 Latest AI
💎 Incredible Automations
🚀 Stuff that actually works
👇 Join the community

[Jack Roberts @JackRob357741]

I tested Claude Fable 5 vs Opus 4.8 on website design. It wasn't even close 😳

🏆 Fable 5 wins 71% of head-to-head design battles
🎨 One prompt built a full product site — crisp, animated, ready to ship
💡 The small details are what separate it — spacing, fonts, motion

Full comparison here 👇
https://t.co/9LNxu9VHoS

[Jack Roberts @JackRob357741]

You can now build a full design system inside Claude and push it straight to Figma, PowerPoint, Miro or PDF 🚀

🎨 Drop in a GitHub repo or image for instant brand styling
📊 Auto-validates your design as it builds
💰 Use GLM 5.2 at 1/6th the cost to keep building when you hit limits

Full walkthrough here 👇

https://t.co/Q6IpKlkTmF

[Jack Roberts @JackRob357741]

Comment "LEARN" and I'll send you the setup guide.

Hermes just learned to repeat your work.

Now it never has to be told twice.

Think of your most boring task.

Same clicks.

Same order.

Every time.

Now show Hermes once.

It saves the whole thing.

Next round, it just runs.

Hands-off, start to finish.

It even double-checks itself.

Like training a new hire once.

Then they never forget.

That boring task isn't yours anymore.

Hermes owns it now.

Go automate the next one.

#hermes #aiautomation #aiagent #workflow #productivity

[Jack Roberts @JackRob357741]

You can now literally talk to your Hermes agent like it's your co-founder 🎙️

💬 iMessage + WhatsApp integration — chat where you already work
🔁 Background agents keep building & researching while you sleep
🗓️ Schedule automations in plain English — no more cron nonsense

Watch the full walkthrough 👇
https://t.co/DDU30sE7w1

[Jack Roberts @JackRob357741]

Imagine one operating system that connects every AI tool you use 🚀

⚡ Claude Code + Hermes + your whole stack — no more silos
🌙 Runs research, builds & reports while you're asleep
📊 Dashboard showing everything your agent has ever built for you

This is level 7. Watch how to get there 👇
https://t.co/Y72KCdILI3

[Jack Roberts @JackRob357741]

Comment "M3" and I'll send you the link.

MiniMax M3 just outscored GPT-5.5 on benchmarks.

At 5% of the cost.

This is what sparse attention does.

M3 only attends to the most relevant tokens.

Not everything at once.

Handles a million tokens.

Twenty times less compute than standard attention.

Open weight too.

Run it yourself.

Or via API at $0.30 per million tokens.

GPT-5.5 costs multiples of that.

Like Michelin-star quality at fast food prices.

Frontier performance.

Fraction of the cost.

The expensive models just lost their main argument.

#minimaxm3 #sparseattention #openweight #aicoding #aimodels

[Jack Roberts @JackRob357741]

Stop using one model when you can have a council of AI experts working for you 🤯

🎯 Opus 4.8 = the orchestrator calling the shots
💰 DeepSeek V4 Pro = 87 cents per million tokens doing the heavy lifting
🏆 GLM 5.2 = the wildcard that rivals Opus at a fraction of the cost

Here's exactly how to build this 👇
https://t.co/DDU30sE7w1

[Jack Roberts @JackRob357741]

Most people don't know you can escape Claude Design's limits entirely 👀

🛠️ Download your build as a zip
🌐 Run it locally with Open Design
🤖 Power it with ANY model you want — Claude, GPT, GLM

Never get locked out again. Watch this 👇

https://t.co/8OMjbj7IGX

[Jack Roberts @JackRob357741]

Hermes just dropped its biggest update yet — and it's faster, smarter AND cheape🔥

🧠 Ministry of Experts: multiple models debate, one gives the final answer
⚡ Cache system means the whole panel costs pennies to run
📊 Claims 8% better than Opus 4.8 solo — and I believe it

Full breakdown here 👇
https://t.co/DDU30sE7w1

[Jack Roberts @JackRob357741]

The biggest Hermes Agent mistake I see? People never give it a soul 💡

📋 Soul.md = your goals, business, how you work
🔁 Two people, same question → radically different answers
📈 The more it knows you, the more powerful it gets

Here's how to set it up properly 👇
https://t.co/Y72KCdILI3

[Jack Roberts @JackRob357741]

What if your AI ran 100% locally — offline, private, $0 per token, forever? 🤯

🛡️ No rate limits, no region locks, no subscriptions
🧠 Qwen 3 locally = last year's frontier model quality
🏗️ Build your own NotebookLM clone and own the whole stack

Full setup walkthrough here 👇

https://t.co/q6PFHz2nDT

[Jack Roberts @JackRob357741]

Stop using one model for everything inside your AI agent. Here's my new setup 🧠

🟣 GLM + Opus = big brain debates & deep research
🟣 DeepSeek = high volume, low cost tasks
🟣 Hermes routes to the right model automatically

Build this system yourself 👇

https://t.co/ur8AQ9MYTV

[Jack Roberts @JackRob357741]

Claude Fable 5 got banned with zero warning. Had it for 72 hours. Gone. 🚨

⚠️ You don't own the models you rely on
🔒 Any region, any time — they can pull the plug
💻 Here's how to run world-class AI that can NEVER be taken from you

Do this right now 👇

https://t.co/aGSce039cA

[Jack Roberts @JackRob357741]

Claude Design 2.0 just dropped and it changes everything 🔥

⚡ Build websites, apps & pitch decks in minutes
💡 Credits finally go further — no more bleeding tokens
🎯 Edit directly on the canvas in real time

Here's the full 5-level breakdown 👇

https://t.co/8OMjbj7IGX

[Jack Roberts @JackRob357741]

Everyone's saying Fugu just killed Claude Fable 5. I tested it so you don't have to 👀

🟠 It's not a model — it's a router pretending to be one
🟠 Took TWICE as long as Opus on the same tasks
🟠 The real surprise? GLM 5.2 at 1/6th the price

Full breakdown here 👇

https://t.co/FmrCBx8DJy

[Jack Roberts @JackRob357741]

You're overpaying for the wrong brain. That's the one system most Hermes users completely ignore.

🔥 MiniMax M3 matches GPT 5.5 on coding benchmarks — at 4% of the price
⚡ 1 million context window, live web browsing, and it sees images AND videos
🎯 Hermes is the car. The model is just the engine. Swap it anytime.

Stop being loyal to the wrong model 👇

https://t.co/n0t3DpzBZo

[Jack Roberts @JackRob357741]

Vibe-coded sites look like slop for a reason. Claude doesn't have visual taste by default.

The Design Skill fixes that. Claude learns to make visual decisions first. Before it writes a single line.

Then add 21st Dev via MCP. Real components. Real design decisions. Not slop.

Like giving your builder a style guide. Before they build anything.

One setup. Claude stops generating slop. Starts building with taste.

#claudecode #21stdev #vibecoding #webdesign #github

[Jack Roberts @JackRob357741]

Asked Hermes to find 6 contacts at OpenAI & Anthropic, research them, write personalised outreach, and build an HTML report.

🟠 Clay MCP connected in under 2 mins
🔵 Verified emails pulled from 150+ data sources
🟣 All from a single voice prompt inside Hermes agent

This is what the research stack looks like 👇
https://t.co/qphhwVEBkK

[Jack Roberts @JackRob357741]

Weekly reports eat time you can't get back. Agent Studio runs them while you sleep.

Show it your existing workflow. Spreadsheets, reports — whatever you already built.

It reads your logic. Keeps your rules. Converts it to an agent that runs itself.

Outputs directly to Slack or Teams.

Like cloning how you already work. Then automating it.

You don't rebuild the process. You hand it over. It runs itself.

#aiagents #automation #aiworkflow #slack #microsoftteams

[Jack Roberts @JackRob357741]

Don't just use an AI script. That's the AI slop apocalypse. People see it coming a million miles away.

🧠 Use Claude to find outliers, deconstruct what's working, and do the deep research
✍️ Then YOU write the script like a human with a brain
🎬 Let Claude handle the editing, images, timestamps, and audio enhancement

The strategy is yours. The grunt work is Claude's. 👇

https://t.co/PuBIbAOuJG