Technical Email Marketer|| Social media marketer|| Lead Ambassador|| Researcher|| Content writer|| Space host
Joined March 2024
- Tweets13.8K
- Following1.6K
- Followers1.4K
- Likes9.7K
Prof Ray retweeted
Rill is built around a useful idea which is turning whatever you’re reading in the browser into work for Claude Code or Codex, without constantly switching apps.
From the product walkthrough, the strongest part is how it connects pages, existing projects and agent tasks in one place.
One thing I’d improve is support for more agents, especially Cursor. That would make the workflow useful to a wider range of developers.
I’d also want clear controls over what page context gets shared and when, particularly for tasks using logged-in sessions.
The concept is compelling. The real test is how reliably it handles everyday work.
Prof Ray retweeted
-> Claude Opus 5.5 $4/$20
-> Claude Fable 5.1 $10/$20
-> GPT-6 Astra $2 / $10
-> GPT-6.1 Sol $2 / $10
-> Claude Sonnet 5.5 $2 / $10
-> Gemini 4 Argon $2 / $10
That list up here is the typical API pricing per million tokens of using the current AI Frontier models for Input / Output respectively. These are the barriers a lot of people don't consider when talking about building with AI.
-> "Cost has been one of the major limitations of working with AI."
But with @say_gm_, you have a great alternative that works very well especially these times when everyone basically uses or builds or using with AI
Essentially, SayGM is an LLM gateway built on Bittensor (Subnet 28) that gives developers, AI power users, and tools drop-in access to top-tier AI models (Claude, GPT, Gemini, and open-source models all included)
And there's a lotttt to benefit from by using SayGM. One of the most persuading one is that instead of creating separate accounts and managing billing across several models, you can just directly point your existing code or apps (like Hermes Agent, OpenClaw, or Open WebUI) directly to SayGM.
IT IS IMPORTANT TO NOTE THAT ALL THESE HAPPEN AT OR BELOW THE PRICE. (I attached a screenshot of the discounted price of the models on the dashboard included in this tweet)
Another thing is that while most AI providers promise data privacy through terms of service ("privacy by policy"), SayGM runs inside an Intel TDX Trusted Execution Environment (TEE) so true Privacy is guaranteed.
This is exactly the type of tool you will envy to have. (My personal ref ink is in the comments). Use it to join
Prof Ray retweeted
SayGM is addressing a specific problem for Bittensor subnets: how to incentivize customer acquisition, not just service provision.
Through Proof of Customer, Subnet 28 directs part of its emissions to an on-chain affiliate pool. Affiliates earn a share based on the qualifying revenue generated by customers they refer during their first 30 days.
The reward is capped at $1 of alpha per $1 of qualifying revenue, valued at the round's close. If referred revenue exceeds the available pool, the rate drops.
@say_gm_ is tying affiliate rewards to actual customer spending, not just clicks or signups.
Bittensor subnet 28 (@say_gm_'s) affiliate pool raises an interesting question about how Bittensor subnets allocate incentives.
What happens when emissions reward not just the people supplying a service, but also the people bringing paying customers to it?
That's the idea behind Proof of Customer on SN28.
Affiliates share a pool funded by subnet emissions, with allocations tied to qualifying customer revenue.
The reward is capped at $1 of alpha per $1 of qualifying revenue, but the actual rate can be lower when referred revenue exceeds the pool.
The distinction matters because signups alone aren't the outcome being rewarded.
Prof Ray retweeted
bittensor:native 🌳👽🦉 @say_gm_
Prof Ray retweeted
I used @ViggleAI to animate the TestDrive car, and the output came out clean.
One thing I’d improve is template discovery. I had to scroll through the trending section to find a motion that worked for the car.
Categories for mascots, product characters and brand animations could make that process quicker for creators.
Good output, but finding the right starting point matters too.
Prof Ray retweeted
Bittensor subnet 28 (@say_gm_'s) affiliate pool raises an interesting question about how Bittensor subnets allocate incentives.
What happens when emissions reward not just the people supplying a service, but also the people bringing paying customers to it?
That's the idea behind Proof of Customer on SN28.
Affiliates share a pool funded by subnet emissions, with allocations tied to qualifying customer revenue.
The reward is capped at $1 of alpha per $1 of qualifying revenue, but the actual rate can be lower when referred revenue exceeds the pool.
The distinction matters because signups alone aren't the outcome being rewarded.
Prof Ray retweeted
POV: My mood ever since I became a tester on @testdriveso 🚗🥰
I animated the TestDrive Car using @ViggleAI, and this was fun to try!
I uploaded the car PNG, picked a motion template, and watched it come to life.
The process was simple, and I like how the templates make it easy to animate images without complicated prompts.
My honest take?
It's a fun tool for creators who want to turn static images into entertaining videos.
Frens, go give it a try...
🤝 Paid partnership
Prof Ray retweeted
I gave @OnepinAI a sentence packed with unfamiliar names, prices and dates to see how its voiceover handled them.
It pronounced the wording correctly, including the names and figures. That matters when a voiceover needs to communicate details accurately, not just sound smooth.
One thing I’d like to see is more control over how specific names are pronounced.
Worth a test if you work with voiceovers.
Prof Ray retweeted
An AI analyst is only useful if you can examine the reasoning behind its output.
@alpharidge_ai lets you ask its AI analyst to build stock screens, scripts and alerts, while showing the platform data used for each answer.
That visibility matters in a domain where an answer can sound convincing without being well supported.
You still need to evaluate the output yourself, but being able to inspect the underlying data gives you something concrete to work with.
Researching a stock and turning that research into a working strategy are two different tasks.
Most trading tools help with one part of the process.
@alpharidge_ai brings research, screening, scripting, backtesting and automation into one platform.
You can even bring existing TradingView Pine scripts into RidgeScript rather than rebuilding your work from scratch.
Its AI analyst can also build screens, scripts and alerts, while showing the data behind its answers so users can verify the output.
The goal is to move from studying the market to putting your own rules into practice, without constantly switching between tools.
Prof Ray retweeted
Handles Mentioned:
AlphaRidge: @alpharidge_ai
Team Rizzo: @TeamRizzoAI
Frank Rizzo: @FrankRizz07
xitter.cf/LakePtr/status/2110271…
Prof Ray retweeted
Finding the trick is one job. Handing the company something they can use is the other.
@perturbaix gives them a number for how easy the model is to fool. A picture of where it’s weak. The actual tricks that worked, so the team can teach the model not to fall for them again. And a receipt on the chain so nobody has to take their word for it.
They want that for places that trust a photo. A bank checking a face. An insurer looking at a damage shot. Anyone about to ship a model and hoping nobody’s already found the gap. Presales and first revenue are the target, not a done deal.
Image models are what’s running. Text and chat come later.
perturbai.io
🤝 Paid partnership
You look at a photo. You see a stop sign. An AI looks at the same photo and can be sure it’s a speed limit sign, because someone changed a few pixels you can’t see.
@perturbaix is Bittensor Subnet 26. People get paid to find those invisible changes before someone else uses them. Someone else checks that the model really flipped and that you still can’t tell. Best finds get paid. Copies don’t.
What’s live is testing on image models. It’s been running since May. Text and chat models are later, not now.
perturbai.io
Prof Ray retweeted
Introducing Mailcheap, powered by Migma. Open source.
@marclou open-sourced Mailcheap: newsletters at AWS prices, on your own AWS.
Great idea. But you still need SES approval, MongoDB, Redis and a server to run it.
So we hosted it: mailcheap.migma.ai
1. Sign in with Migma
2. Add your website and domain
3. Write your issue, or let Migma design it in your brand
4. Send
We handle sending, warmup, compliance and design.
Same price as AWS. No AWS, Mongo, Redis or maintenance.
Code: github.com/MigmaAI/mailcheap…
credit: @marclou
So many people asked me to make it open-source, so here you go:
MailCheap — Self-hosted newsletters on Amazon SES:
💸 $0.10 per 1,000 emails
✍️ Editor with live preview
📰 Multiple newsletters
📝 Signup forms + API
🌐 Public archive
📊 Stats & growth charts
🧹 List cleanup
📬 Bounce + spam protection
→ github.com/marclou/mailcheap
92% cheaper than Beehiiv (for me).
Have fun!
Prof Ray retweeted
Gemini 4 Argon is now live in MigmaAI.
Brilliant email designs, compatible with every email client, powered by Zinn.
Try it on your brand today.
Prof Ray retweeted
I built a Web3 research agent with @simdotai to turn project research into a concise brief.
The first output was around 1,300 words.
After rebuilding the workflow, the revised brief came down to 684 words, with conflicting figures flagged and unconfirmed claims labelled instead of presented as facts.
One thing I’d improve: a clearer way to verify sources beyond search snippets. The agent retrieved 22 links, but only opened 5 in full.
This is a useful workflow, but source quality still needs human attention.
