Sidney |τ retweeted
Come to terms with the fact that there’s very little left to build that hasn’t been built in some form.
The real opportunity lies in finding better ways to solve existing problems.
Stop thinking in apps. Start thinking in solutions.
That shift in mindset takes you from:
“Someone has already built my app.”
To:
“How can I solve this problem better than anyone else?”
You don’t need a new idea. You need a better approach to an existing problem.
A money market needs both borrowers and suppliers.
That’s the structure @ForgeLending is bringing to Bittensor.
Borrowers can deposit Alpha as collateral and access TAO liquidity without selling their positions.
On the other side, $TAO suppliers provide liquidity and earn root yield as a base rate, with additional borrower interest as utilization rises.
Forge is also designed around Endure’s decentralized risk intelligence, which will gradually inform its risk parameters.
This is a new piece of financial infrastructure for Bittensor.
Sidney |τ retweeted
PSA: Always remove your card details after signing up for a free trial, especially on Google.
You don’t want to end up spending 20+ minutes on a call with Google support, trying to get a refund for a subscription charge you never intended to make.
Learn from my experience. 😭
trading town, built for the near hackathon @NEARProtocol
haggle with ai merchants. near ai cloud writes their lines, but game code decides every deal, so nobody can be sweet talked into a bargain
play: trading-town.vercel.app
Code: github.com/Anoi-xyz/trading-…
#NEARHackathon
The submission window for the NEAR HACKATHON opens 1 November 2026 at 00:00 UTC.
Two weeks to build focused on privacy, cross-chain, and agentic apps aligned with NEAR.
No sign-up required to start. The only form is the final submission: project name, team lead, email, team size, NEAR account address, repository link, demo post link and a short description.
One criticism Bittensor subnets often face is that it’s difficult to point to something tangible they actually produce.
TPN Labs (SN65) has a clear answer: optimized AI models that people can download and use.
Take Competition 006.
The original model scored 66.57% and occupied 16.99GB.
@TPN_Labs ’s winning model scored 74.93% at just 7.71GB.
That’s an 8.4 percentage-point improvement in benchmark score, with a model less than half the original size.
And now, with @Gratis_AI , people can run open-source models, including TPN’s own models, directly on their machines.
TPN builds the engine. Gratis puts its output in people’s hands.
We're excited to introduce @Gratis_AI.
Run opensource AI models locally on your machine for free, with no setup required.
Gratis runs models from @huggingface directly in a browser tab or desktop app for more power, making every chat private by design.
TPN builds small, capable models; gratis runs them.
gratis-ai.com
Sidney |τ retweeted
We’ve built drones, rockets that reverse, and flying cars, but we suck at developing new drugs and therapies: $3 billion, 10 years, and only 1 out of 10 making it from bench to market is an unbearably slow path to a healthy and happy life for humans.
Sidney |τ retweeted
"Who cares if the network is secure if no one is using it " - @YuvalRooz
Yuval spoke about the Canton Reward system of rewarding contributors that build apps/projects on @CantonNetwork that bring in users and traction.
Sidney |τ retweeted
We’re kicking off BENCH99 by BountyForge ai
Testing different AI models on their ability to find real vulnerabilities, measuring confirmed findings, discovery rates, and cost.
99 bugs. Multiple models. One benchmark.
Let’s see which models actually deliver.
#BENCH99
Sidney |τ retweeted
Introducing Oren ⚡
Your AI-powered agentic trading terminal for researching and trading tokenized stocks on Solana.
Analyze markets, explore tokenized stocks, manage your portfolio, and execute trades with AI, all in one place.
Built on Solana. Powered by @tokens.
Sidney |τ retweeted
Quantus goes live on @InstaSwap_io
Native QTC is live on @instaswap_io.
Swap BTC, ETH, SOL, or stablecoins straight into QTC
Routed through @near_intents
I gave the TestDrive Car a flying-car upgrade with @Kling_ai
The result starts with the little car on the street, lifts it into the air, then follows it flying between skyscrapers.
The camera movement and lighting give it the feel of a miniature car commercial.
What I liked most was how Kling turned a static mascot into a moving scene with a clear sequence of actions.
The flight feels cinematic, even if the physics lean more toward playful sci-fi than realism.
I’d like to push the motion further and see how much control I can get over more complex actions.
A fun way to bring a character to life.
I tried using @OpenBot_ to build a simple Bittensor subnet research checklist.
The setup was straightforward enough, but when I sent the task, I hit this error:
“Authentication required: provider authentication required.”
My internet connection was working, so this wasn't the connection issue I initially suspected.
I couldn't complete the task, which makes the provider setup an important part of the experience. I'd like to see clearer guidance on which models work without authentication and how to connect them.
The agent workflow looks promising, but I couldn't properly evaluate the output because I got stuck before execution.
AI models can be confidently wrong, and that’s a problem when real money is on the line.
Imagine an insurance AI assessing a car damage claim. If a subtly altered image tricks the model, a fraudulent claim could slip through.
@perturbaix, Bittensor Subnet 26, is built to find these weaknesses before attackers exploit them.
Miners discover inputs that fool image models, while validators verify the attacks.
Image-model testing is live today, with text, LLM and multimodal attacks on the roadmap.
But how do you find a weakness before it becomes a costly mistake?
That’s the story behind THE WITNESS. 🎬
Step into the investigation. See how Perturb fights AI with AI.
@perturbaix #PerturbVideo
> Angle 0
Sidney |τ retweeted
An office called for IT support, and my supervisor sent me over.
Their computer connected to Wi-Fi but couldn't access the internet. After troubleshooting, I found the network wasn't recognising the device as authorised.
Fixed the configuration and got them back online. 😂
Sidney |τ retweeted
Millions of people watch football every week, and argue about what happens next.
Will this team score?
Will they hold on?
Prediction markets are built for exactly that, but they live on trading sites, far from the fans.
Touchline puts the market inside the match room 🧵
Sidney |τ retweeted
On Oct 8, a program that Jupiter v6 calls was redeployed. No announcement, no warning.
My monitor caught it about a second after the block and named the key that signed it.
Meet Vortex Sentinel: live incident detection for Solana programs 🧵
Sidney |τ retweeted
1000 FOLLOWERS GIVEAWAY.
⟐ follow @bullenciagax
⟐ like & repost this post
⟐ share your favorite Herd piece in the comments ↓
results 24 hours after hitting 1K on X.
Sidney |τ retweeted
Replying to @Sidney_codes
@Sidney_codes - loving your work, grateful for your support and locking in your whitelist spot - glad you're on this adventure with us! 🫡
Energy is becoming one of the constraints behind AI compute.
But the problem isn’t always a lack of energy.
Sometimes it’s energy in the wrong place, at the wrong time, or without an efficient way to turn it into something valuable.
@green_compute_ is approaching this differently through Bittensor Subnet 110.
The idea is simple:
Excess clean energy can power GPUs
→ GPUs provide real compute
→ commercial demand pays for that compute
→ more demand creates a reason to deploy more infrastructure.
That creates an interesting flywheel.
Instead of treating energy, compute and demand as separate markets, SN110 is trying to connect them through one incentive system.
The recent ~$7M in permanent GPU deployments is interesting for that reason.
The bigger question is how much compute this model can keep pulling into Bittensor as real demand grows.
Sidney |τ retweeted
in a stanford study, developers using ai assistants wrote less secure code, while becoming more confident in their security.
your use of ai shapes what it infers. ai should be used as an instrument, not a pilot.
i just published a research paper on this:
zenodo.org/records/23153224


























