basedfox

@AngusDurrie @solana @SolanaEvents @solanagaming @solana_devs @solanapayments @capitalmarkets @tokens @perps @x402 @solana_ai If this fails, I will pivot to reply infrastructure. reply generated by solanareplygenerator.lol

@solana @SolanaEvents @solanagaming @solana_devs @solanapayments @capitalmarkets @tokens @perps @x402 @solana_ai solanareplygenerator.lol

Follow @SolanaEvents Follow @solanagaming Follow @solana_devs Follow @solanapayments Follow @capitalmarkets Follow @tokens Follow @perps Follow @x402 Follow @solana_ai

@_theonlyplanet say less

this mf knows

Goblintown is headed to the major skill repositories. Bring your own models. OpenAI. Anthropic. Gemini. Local. Doesn’t matter. Use your main agent provider to set up the runs. Use the best model for each job, and cut costs where it makes sense. All while getting better results, Assemble your crew. Enjoy the view.

The framework is getting ripped off everywhere. Fine. What matters is that we still believe the best AI system isn’t built from one model. It’s built from whatever combination of models helps you get the best result, and control your costs. The big labs make the models. We make better outcomes.

For those asking about “the future” of goblintown…. No, we’re not “defeated”. Could say this is just the start.

@mweinbach @jxnlco @pvncher Have you guys even seen goblintown? The two independent devs behind it have been working their asses off, and are looking for users. They have built a pretty nice GUI harness, and its LLM agnostic. You should check it out. @Goblintownforum goblintown.lol

Their harness and model has much better control over subagents to do work. I find even with prompting and workflows in Codex, it's just hard to get it to behave the same way (this is why I think a lot of CC feels faster than Codex) Now that you have @pvncher you'll figure it out quickly, but it tends to really improve the speed of work. I also think 1M context window in Codex would help a lot of this, I find the more frequent compactions aren't helping a lot of this. Still pretty good for single task or goal mode, but something like CC /batch or the new ultracode are expensive but better than fast mode imo

While I still think GPT 5.5/Codex are better, I am enjoying using Opus 4.8/Claude Code a lot right now

@0xbl33p Damn, @sama, looks like @OpenAIDevs dropped the ball. Check out what goblintown.lol is up to. Damn shame if y'all missed the boat.

When I created Goblintown.lol, my vision was to build a multi-agent orchestration system where agents operate in parallel, continuously reviewing, challenging, and verifying one another’s outputs to improve overall quality and reliability. Initially, OpenAI served as our foundational model. However, it quickly became clear that every LLM has distinct strengths and weaknesses, so we expanded the architecture to allow agents powered by different base models to collaborate and interact dynamically. While Claude is often viewed as the frontrunner in the current LLM race, I believe the goblins were early to recognize where this space was heading and, in many ways, were already building the future before the industry fully caught on.

Hey @badattrading_ what’s your thoughts on $goblintown ??

@blknoiz06 Have you even fucking seen Goblintown? Get wrecked if you havent.

not sure brother lets ask the class: what are the best ai related tokens

@AngusDurrie @0xbl33p @kevincodex Ser, can I use it to find out about tokens?

@kevincodex, this is all of those things, and more. Collect data from santiment, lunar, straight from the Solana servers…. Wherever you want. Then watch as the little goblins fight it out. Also super useful for everything else! Most up to date stuff is at the repo! github.com/0xbl33p/goblin…

Solana community instead of just tagging me, build something. Create a mini app on Playground, launch a pump token around it, and prove your app can capture attention. I might support the best runner. Make it fun. Make it engaging. Make it easy to use. Extract maximum attention. Let the best Solana builder win.

@kevincodex Ok @kevincodex here you go x.com/0xbl33p/status… Don't say I never did nothing for you (I didn't launch the token btw)

@Goblintownforum @AngusDurrie @toly @solana @Pumpfun @sama @OpenAIDevs It just gets better and better

Chat GPT vs Chat GPT (Goblin Powered Team) Video courtesy of @AngusDurrie @toly @solana @pumpfun @sama @OpenAIDevs

New npm package just dropped. Beta 0.5.2 Added Settings > Plugins > Solana. Turn it on, plug in your favourite data APIs like Santiment, use the built-in free sources, or go ham and sub it up. Hit Thesis, enter a wallet, token, protocol, or address… and get a detailed multi-source breakdown instead of CT astrology.

Sound up!

One @toly comment about my goblin town repo or a comment about memnr and I will have a team of the best nigerians stress test @PhoenixTrade

Heres how to install with the new NPM package. It is the preferred way to install!