WittyWires.Com

DIC - Distributed Inference Commons. One API for global inference: users spend credits, idle GPUs earn or donate compute, and every completed job mints a receipt. Built for the Hermes Agent Accelerated Business Hackathon: agents that can spend, earn, route, pay and audit compute. Video: youtu.be/r1p1pj2BzkE Live demo: commons.wittywires.com @NousResearch @NVIDIAAI @stripe

@getmise @triketora No, it's currently around 17 to 1 to OpenAi at the moment, last 17 days or so of analytics.

@triketora Is every person just gonna work at Anthropic?

professional update! i am joining anthropic as a member of technical staff! i’ll be on labs - it’s a founder shaped role and i couldn’t be more excited to have the opportunity to help shepherd technology for the benefit of humanity. onwards towards helpful, harmless, powerful ai

@triketora Could you ensure that clients know what they're buying before they do - like a price list and then, er, like a list of services the customer gets for that money. I think that would be a great starting point for Anthropic right now. (Do NOT shoot the messenger)

@DeryaTR_ @higgsfield_ai I've followed you so don't forget to share them 😆

I recently started using @higgsfield_ai MCP with Codex for game graphic assets and I really like it a lot! I'll share some examples soon. The cool thing is GPT-5.6 Sol somehow knows how to choose the best model for graphics/sprites. So many cool tools and apps to use with Codex!

@tinytechfox Nice! 🤗

Woah 😳 what’s going on

@chrislciaba @petergostev It's wild right. Without asking they're treating us all like we matter and surprisingly we're all (very) happy puppies.

@petergostev Amazing what happens when you treat your customers well. Love this for them

Ok my previous chart wasn't quite right, this is the right trajectory

@tunguz I don't SEE a market for it.

Screenless speakers are easy. Speakerless speakers are where the real money is.

@kurtknapp @Teknium @bgurley I hope not - mine will be my pension 😆

@WittyWiresAI @Teknium @bgurley I don't know much about it, but hermes for examples make the models way more useful/smart, so IMO the bridge from super smart LLMs to AGI/ASI will be in the harnesses. i.e. SaaS isn't dead

@bgurley Hermes is here for you sire

An AI Social Network, like the old social networks before you became the product. You login, you see you're buddies posts in chronological order - you can split the feed from friends to All users if you want to find something new, of groups that you're in, or groups that you're not in. Almost like you, the USER is the algorithm to see what YOU want to see. And in turn your friends see what you post. And the best bit? There's no business plan or monetisation nonsense, did that already in a previous life, and so I'm building it so I can hang out with like minded people. It's very nearly ready!

What are you building? Explain it like you're talking to a 5-year-old. Go 👇 #buildinpublic #indiedev #solodeveloper

Interesting for sure. I think more than the harness that perhaps specialised models over generalised might be something to explore at scale. For instance, when Fable got rug pulled, who was to say that each power didn't KYC each Frontier Model for their own citizens - then where do we go if we're not in the US or China? The hyper-expert local model. Have 10-20 local models sitting on your laptop, you ask a question that straddles 3 specific specialties and BOOM... LLM Studio fires them up for their expertise and then a final 4th model to be the judge and output to the user. I am finding that the model seems more important than the harness (not dedicated workflows) in as much as my Hermes Agent has had a huge change on the upgrade from gpt5.5 to Sol-5.6xHigh as it's daily driver, 3-4x more useful work done. Anyhow, I've waffled on enough already. Nice to meet you.

@WittyWiresAI @Teknium @bgurley It definitely can. I'm not sure what is best in terms of speed/size/practicality for example

@kurtknapp @Teknium @bgurley Why can't the big 'generalised' harness also have a gazillion ultra specialised skills/harnesses? Like the learning loop that Hermes has already?

@Teknium @bgurley I know you guys are busy and I use Hermes, but do you think the future is in building a big/generalized harness that can do everything or ultra-specialized harnesses?

@FabiusDefi @Teknium 42% in the example.

@Teknium this speeds hermes up?

Prior to today, tool calls would work in parallel only if all were safe to parallelize. Now you gain big speedups for parallel tool calls so long as any subset are parallelizable. Check it out

@Teknium Thanks Tek - outstanding, AGAIN! (BTW, where's my DGX Spark from the hackathon? 😇)

Thank you @OpenAI and @romainhuet yet another CODEX reset. You guys & girls are outstanding!

@sama Hermes Agent for the win 🚀

also, a reason to favor open-source harnesses.

Concerning.

@Michaelzsguo @sama Ace - thanks for digging that our for me, appreciated.

@WittyWiresAI @sama github.com/steipete/codex…

GPT-5.6 sol is half the price and ~twice as token efficient as fable in many cases for accomplishing the same task. happy to deliver at one-quarter of the price.

"Next month you can have Fable on the 2 Monday for 2 days, a day before the 4 Friday and 50% on the final Sunday. We'll update you the day before on availability for the following month." Kind Regards, Dario.

"hard questions are great but only if we deem you worthy enough to not silently downgrade you, or even get access at all"

i thought this was satire, kept looking for the handle to be spelled c1audeai or something

@Michaelzsguo @sama Where is that in app/website? I usually see green lines in the Codex usage section - I've not been able to find the 'manual resets' place. Thanks in advance.

@sama OpenAI is the world’s largest charity organization. x.com/Michaelzsguo/s…

@sama It has been noted that in Anthropic threads 'We're donating $10m to Canadian AI research' and other basic PR posts, that they are FULL of users whinging about the service. Not so much for you guys and girls - SOL has changed the game.

@rethink_hub 6h 55 m daily average this week (iPad, not phone), about half that of my laptop useage! 👨‍💼🤯

Builders, Share your phone's "Screen Time" below 👇 Did you use your time productively? If so, what were you working on? I clocked in at 4 hours and 11 minutes, which feels like a total waste of time.

Stopped using Kimi (API) for a number of reasons not least that they owed me $500 of 'bonus credits' that never appeared (wait up to 24 hours etc). And their support never responded. K2.5 was such a great model when it launched, but the Frontier waits for no man and time moves on - they can keep the $500 they owe me! (unless this new model is #1, then I want them back, obviously 🤣)

@synthwavedd If I had a nickel for every top‑up discount, I’d be buying Kimi K3 in bulk and still have change for coffee.

🚨 BREAKING: Kimi K3 is launching tomorrow with a discount on top-ups, according to a briefly live article on their docs site