Product @boundlessHQ l My own thoughts that will probably change
New York City
- Tweets 4,170
- Following 1,037
- Followers 3,416
- Likes 7,030
In OKC this weekend for a buddy’s wedding, catching up with 13 friends from engineering school. They now work at engineering firms and oil & gas companies.
0 use Codex or Claude Code.
1 pays for ChatGPT, mostly for sports news.
Day 0
This is my thesis: no one uses AI. I repeat, absolutely no one. We live in a bubble.
Even among my friends who pay for it, when I ask them to open ChatGPT and show me their queries, it’s the same handful of basic things.
Most don’t even know they can upload a photo and ask questions about it. Connecting Gmail so an agent can read and send emails blows their minds. An agent opening a browser and checking them into a flight? They’ve never even heard of it.
The massive challenge right now is adoption, and then getting people who already signed up to actually use what they’re paying for. Most have absolutely zero clue what’s possible.
Imagine the compute shortage when everyone starts using AI like the top 1% of users do today.
4
5
548
Jacob retweeted
The Laya decision model is free on AI Gateway via @BoundlessHQ through Oct 31.
Use it to route agent work, triage support, and check guardrails.
𝚌𝚘𝚗𝚟𝚊𝚒𝚒𝚗𝚗𝚘𝚟𝚊𝚝𝚒𝚘𝚗𝚜/𝚕𝚊𝚢𝚊
vercel.com/changelog/laya-de…
7
11
73
4,920
Decision models are an incredible tool for high-volume workflows where cost and latency matter and the task is a narrow decision.
Excited to serve Laya free for a limited time through @vercel AI Gateway. Try it out of the box, or fine-tune its open weights for your own use case
Laya is now available on @vercel, served by Boundless and free to use.
An open-weight decision model for fast, structured decisions in agents and workflows.
2
1
14
1,076
One line command to add @BoundlessHQ to any harness or desktop app
in a space where everybody is asking you to come to their new harness
Boundless comes to your harness instead
Try it today, this is the best thing I've worked on this year
1
1
4
997
Hottest thing in inference right now: “we serve model X at Y tok/s.”
Yeah cool bud, at what concurrency, on how many GPUs, and at what cost per million tokens?
Y tok/s != we are a good provider
Introducing Prism
TL;DR: Prism is an inference cloud for open source LLMs. We use agents to optimize our deployments for cost, latency, and throughput (we serve DeepSeek V4.1 Flash at 547 tok/s).
prisminference.com
3
865
Setup Boundless inference with one command from the terminal
If you want to setup @BoundlessHQ as a provider for your harness, all you need is to tell your agent
> Setup boundless.md
I'm convinced this is the future
2
350
RL + OS models is the enterprise stack of the future
Harvey’s AI costs got so high its gross margins plunged from 50% to -50% in 6 months.
Its response: build its own model using open-weight AI.
Startups from Abridge to Rogo are following suit to cut costs and reduce their reliance on OpenAI and Anthropic. My latest w/ @nmasc_👇
3
611
Live and scaling up on @vercel AI gateway this past week.
Serving the most popular OS models competitively amongst the best AI infra providers in the world.
Your favorite inference provider has made the @Vercel AI Gateway lineup.
Boundless is serving DeepSeek V4.1 Flash alongside leading inference providers.
→ 0.3s latency
→ $0.20 input
→ $1.00 output
Explore the full model lineup on Vercel AI Gateway.
1
1
8
1,060
Jacob retweeted
Boundless is now live on the @vercel AI Gateway.
Starting with:
→ DeepSeek V4.1 Flash: $0.20 in / $1.00 out
→ GLM 5.3 Flash: $0.12 in / $0.40 out
More to come. Get access below.
12
12
36
8,357
Christmas for any company with Agentic RL needs
We built a time machine for the web.
Introducing Exa Snapshot: an index of 400 billion historical snapshots of webpages that lets you search as if it's the past.
Snapshot is already being used for backtesting prediction models, RL at labs, exploring the pre-AI web, and more.
3
443
A PM who can turn user pain points into evals might soon be worth more than a top 1% ML researcher.
1
15
807
I’ve published the initial work here:
github.com/JacobEverly/long-…
The repository includes:
- The supervisor and evaluation harness
- Results from four model deployments across 18 sealed tasks
- Completion-versus-cost comparisons
- Roughly 172 hours of redacted agent trajectories
- External-verification outcomes
- Reproducible analysis and visualizations
4
76
This reflects how I think these models are best used today: an almost infinite sea of interns that can take on grunt work, as long as you give them the right direction.
Today we're releasing data on models accelerating research at OpenAI.
Recursive self-improvement could be the most important contributor to AI capabilities over the next few years, but by default it will only be seen inside a few frontier AI labs. Being transparent is more urgent than ever, so we can inform the public discussion on whether and how to pace model development. I ask other AI companies to do the same.
openai.com/index/research-ac…
2
8
663
Is instinct just @NousResearch with a different GTM motion and PMs?
My boyfriend and I got on the Jumbotron last night at the US Open and he wanted the footage. I told him impossible but I’ll ask Instinct bc why not. And of course, Instinct got us the video in 18 hours.
Here's what it did (1/5)
2
3
3,161
We are working with teams where inference is core to the product. If that’s you, reach out.
We built an inference platform, but we’re not a neo-cloud selling cheaper tokens. We work hands-on with teams across product and engineering to integrate, tune, and operate models in production. Resulting in better performance and reduced costs.
We’ve been building Boundless AI, the inference partner for AI-native startups.
Boundless helps teams run open-weight models tuned to their product, for better economics on frontier intelligence.
boundless.network
4
1
13
1,616