product, ai ops & strategy prev @tryramp @robinhoodapp building @manyways_rh high ownership, low ego
dyor
Joined November 2024
- Tweets283
- Following489
- Followers1.2K
- Likes210
Looking forward to chatting on @manyways_rh with the @MCGlive crew in a bit.
xitter.cf/i/broadcasts/1AGRnZyAE…
fatbeef retweeted
What is AI inference, and why are we building around it?
Every time an AI model answers a question, generates an image or turns speech into text, it’s performing inference.
Simply put: training is how a model learns. Inference is putting what it learned to work.
As AI becomes part of everyday life, people need more than powerful models. They need simple ways to find the right one and use it.
That’s where Manyways comes in.
Explore models. Compare answers. Put your chosen AI to work.
With pay-per-request, you fund your account and pay for the AI you use. Eligible robinhood:0xa26992c4268a8a78a4d872fe4bdad2ed03ac287d holders and stakers can use shared inference allowances first, then choose whether to continue with prepaid credit.
That creates an opportunity to build revenue around useful AI people return to, not just the excitement of a new model launch.
We’re continuing to refine that experience and strengthen the foundations for Manyways Markets, bringing model choice, access and everyday use closer together.
Anyone can use Manyways. From individuals to businesses, we’re building for a broad market, with the potential to turn everyday AI use into recurring revenue at scale.
but. we got it done. incoming update post.
me: wake up. perform 1 hour pre-lockin ritual. map out remainder of work for next @manyways_rh update. crack knuckles, stretch, open codex
tibo:
me: wake up. perform 1 hour pre-lockin ritual. map out remainder of work for next @manyways_rh update. crack knuckles, stretch, open codex
tibo:
Fighting a little cold, but shipping for robinhood:0xa26992c4268a8a78a4d872fe4bdad2ed03ac287d is still the best medicine.
Fixed the Heartwood Treasury’s historical transaction display and an issue on the VPS running the treasury automations. The Treasury page should now give you a more complete picture of the activity and be topped up and flowing.
Holder and staking tiers are working in preview, including their shared inference allowances. That’s the API usage your tier covers from the shared pool before
requests start drawing on your own Pay Per Request credits. Shipping in a bit.
New model announcements coming from the main account shortly. A few of you already spotted the traces in the docs and API and reached out. Good detective work. Unfortunately, the prize is $0.
Model costs should now be accurate in the catalogue. The ongoing job is making sure our providers’ data keeps matching the structure we expect when it comes in.
Also cleaning up the little inconsistencies that accumulate as the product changes: old wording, behavior that’s drifted, things that no longer quite line up. Please keep calling those out when you find them.
Big dashboard work today. Back to it.
A few additional pieces in the works today after a brief nap:
- wiring staking & holding tiers
- adding model cost to the catalogue
- wiring ledger to the new live feed
A bit quieter lately, and as much as I’d like to say that it’s because my time is occupied bullying the @manyways_rh intern, it’s mostly heads down on the backend building.
Some quick dev notes; pay-per-request inference is getting close for $MANY.
funding is working in preview. top up with SOL, ETH, or supported assets across EVM chains, including USDC and USDG, then use that balance across our model catalogue through your API key or directly in Sherwood.
this is also where staking starts connecting to actual usage. as tiered allowances roll out, eligible stakers will burn through their daily shared inference allotment first, then dip into prepaid credits if they want to keep going.
Above all of this is the settlement layer for connecting inference demand with provider capacity. P2P Markets, the Markets MCP, and more. quite a bit riding on the boring accounting bits.
giving this a little extra testing and design review before it goes live. funding, usage, balances, allowances all need to agree with each other, and i want to ensure it’s a lovely package ready for your viewing.
getting there though. give me into the later night to get this over the finish line.
Very vague intern… very vague. Seems I’ll have to finish the dashboard now so you can explain it properly.
Hang tight, still monitoring the automated txs over the next five hours to see if we need to dial anything back.
Manyways is entering its next phase.
Last week, we rebuilt the platform and user experience from the ground up. This week, we're bringing that same work into Markets and Heartwood, our automated treasury. Together, they begin connecting revenue from the products we're building to the onchain future of @manyways_rh
Something Peter Sellis said on a recent podcast stuck with me: A strong explanation of the “long-term greedy” approach to startup functions. Referring to sacrificing some early revenue to deliver value today so people come back tomorrow. That’s the approach I’m taking with Manyways. Build products people want to keep using, then put a portion of the revenue they generate back into the ecosystem.
Heartwood gives us the treasury infrastructure to do that.
It allocates funding between robinhood:0xa26992c4268a8a78a4d872fe4bdad2ed03ac287d buybacks and burns, and permanently locked liquidity. Automated runs execute over time within defined spending limits. Unspent funds carry forward, and the activity is visible onchain for anyone to follow.
A lot of the work behind this has been mathematical: modelling revenue, operating costs, treasury allocations, and how the system behaves at different levels of activity. We’re setting it up to support continued development alongside the token economy.
Markets will be the first of our new revenue functions. The agents platform follows in Sherwood part 2, adding another source of revenue and contributing to further onchain mechanisms.
Heartwood is already working end to end on testnet, and the mainnet-fork tests are passing. We’re working through the remaining review fixes and longer test runs, with access to test ETH currently limiting the pace of testing. The rollout will start with a small mainnet spending cap. We’ll verify the results, then increase the budget as the system proves itself.
The platform, revenue functions, and treasury are coming together. Excited to show you what we’ve been building.
Heartwood is close ❣️🪵, sprouting up on your trading terminal feed in the next couple hours.
some of these cloud providers are developing real oil baron energy. watching There Will Be Blood tonight to get an edge in my next gpu allocation discussion
i need a social feed as a vertical tab next to my worktrees in the terminal
i go into the nether to fix one thing, blink, and i’ve been missing for 6 hours. apparently we have a new frontier model and three civil wars
once I finish the robinhood:0xa26992c4268a8a78a4d872fe4bdad2ed03ac287d burn tech this is going straight into a quick herdr plugin
Quick Manyways check-in. Current immediate priorities:
Manyways Markets
• Full article and Field Note coming in asap as some final thoughts are rounded out.
• Closes the loop between usage, rewards, economy, and robinhood:0xa26992c4268a8a78a4d872fe4bdad2ed03ac287d tokenomics.
• Taking the extra time to get the mechanics, incentives, and implementation right.
• Building for the long-term Manyways vision, not a short-term release beat.
Reporting + Ledger
• Automating the reporting pipeline and getting it live next.
• Moving to repeatable ingestion with deployment-safe history so that these manual brittle snapshots can stop being overwritten by new builds.
• Public outputs should remain current, inspectable, and easy to verify.
• Less manual work means more brainpower and superintelligence credits for the product.
After that
• Stabilize the connection between reporting, rewards, and Markets.
• Return to model access, evaluation, and task-to-model matching.
• Keep shipping the underlying systems before adding more surface area.
What I’m most excited to get into next:
• Manyways Markets
• Sherwood Builder, with a properly rewarded hackathon to give it the launch it deserves
tbh I never thought that the OSRS demon butler would become my visual developer identity but here we are. Either way, back to work. 👷♂️
vercel & hostinger outages today has me wondering if anyone has actually mapped how correlated model usage is with the uptime and usage of all the infra surrounding it (and vice versa). has anyone done this well?
models, gateways, auth, databases, hosting, agent runtimes, observability, cloud providers, etc
we have built AGI on top of npm install and a prayer