Product Marketing Lead @near_ai | 2x Founder (1 Exit)
🇦🇷 • 🇺🇸
Joined July 2011
- Tweets1.2K
- Following1.1K
- Followers1.3K
- Likes4.9K
ALEJO retweeted
Proving code works with math instead of tests used to cost too much at scale.
NEAR AI is now #1 on Lean Eval v1, Lean's own benchmark of extremely hard formalization problems.
It runs almost entirely on open-weight DeepSeek V4.1 Flash, so it's cheap.
Next: NEAR contracts.
NEAR AI has claimed first place on the Lean Eval v1 leaderboard.
Lean Eval v1 is an active benchmark maintained by Lean FRO that contains extremely hard formalization problems. We reached first place despite joining the competition just four weeks ago, a full month later than other top participants.
NEAR is very heavily investing in formal verification efforts. Our ultimate goal is to formally prove the protocol and all the core contracts, and to provide a practical tool for others to formally verify their contracts in a fast and affordable way.
The system that enabled us to claim first place is almost entirely powered by DeepSeek V4.1 Flash, and thus is (a) extremely cheap, and (b) has no dependencies on third parties because the model is open weight.
When it comes to formally verifying software, the approach boils down to three pillars:
1. Representing the program in a way that can be reasoned about;
2. Expressing the properties that we desire to prove;
3. Coming up with a proof and formalizing it in Lean.
Until recently, step (3) was prohibitively expensive for practical use at scale. With the system that we used for Lean Eval, we now have it solved. Stay tuned for updates on how we apply it in practice to formally verify NEAR contracts.
Really happy to see NEAR AI Cloud live as a provider on @OpenRouter!
OpenRouter is the most transparent arena in AI inference. Same model, dozens of providers, price and speed side by side. You can't market your way to the top there. You have to earn it.
Proud of the team for getting us there, and a special thanks to @Cameron_Dennis_ for leading this one.
NEAR AI Cloud is now a provider on @openrouter.
NEAR AI offers GLM 5.3 Flash and 1M context, under the provider slug near-ai.
Reachable with the key and balance you already have.
ALEJO retweeted
NEAR AI Cloud is now a provider on @openrouter.
NEAR AI offers GLM 5.3 Flash and 1M context, under the provider slug near-ai.
Reachable with the key and balance you already have.
ALEJO retweeted
In case you missed it last week, @near_ai can now be used on @OpenRouter and its really simple to get started.
If you're already on OpenRouter, private inference is a couple of clicks away as of this week due to @near_ai
The frontier-lab models on NEAR AI Cloud are what we refer to as INCOGNITO. Solid pick if you want frontier capability but rather not be tied to your prompts
If you want to protect IP, we run fully verifiable CONFIDENTIAL models like DeepSeek V4 Flash
(GLM 5.3 drops this week)
where have we seen this before?
2020, Amazon's private label team got caught studying sales data from the same third-party sellers competing on its own marketplace. clone the winners, undercut the price.
they told congress they didn't. the WSJ found the receipts anyway. the EU eventually made them stop.
this week: a mathematician spends a year chasing one of math's hardest open problems, using OpenAI's own tools along the way. before he can finish, OpenAI publishes a proof for that same problem, using the same unusual approach he'd quietly been working. he asks the lab a simple question about his own data. doesn't get a straight answer...
not saying it's the same thing (and neither is he). but there's non zero risk that the tools you're building on turning into your biggest competitor. it's already happened once. and when it does, a policy's "trust us" stops being a real answer.
verifiable confidential AI shouldn't be a nice-to-have!
Mathematician Tristan Buckmaster spent the better part of a year getting an AI to help him chase one of math's hardest problems.
Then he had to ask if he'd handed someone else the answer.
A policy page tries to build trust, but falls short. The infrastructure has to.
On NEAR AI Cloud, models run on secure hardware that proves nobody, not the model's maker or us ever sees or trains on your prompts.
You shouldn't have to ask and hope for an honest answer. You should be able to check.
And you can: cloud.near.ai
ALEJO retweeted
Verifiably private GLM‑5.3 Flash is now on @near_ai!
It matches Opus 4.8 on performance yet costs just $0.075/M input tokens and $0.25/M output tokens.
That is 1/67th and 1/100th of Opus pricing 🤯
cloud.near.ai/models
ALEJO retweeted
NEW LAUNCH
Staking for NEAR AI is live.
Pay for confidential inference and always-on agents by staking NEAR. No card. No cloud account. No custodian holding your credentials.
🧵
We've been running this on ourselves for months. Every person at NEAR AI and NEAR Foundation has their own IronClaw agent.
1.0 is out. Go roast it.
“stronger hook” “make it impossible to keep scrolling”
The fastest way to kill a product is overpromising in your marketing
Overpromise → user tries it → gets disappointed → tells everyone → you lose
Best marketing undersells a little. Let the product do the talking.
Tried it out. So the product scans stuff, then it shows Reddit tells me people post about my company, surfaces competitors I already knew about ages ago, wraps it in a dashboard, and wants me to scream “OMG we saved $60k. At this point I’m convinced half of startup hype is just polished demos. How’s this the “ultimate CMO” for me
One prompt. 1 minute. 8 GB back on my MacBook.
Claude went through every installed app, flagged 12 I didn't need, and safely uninstalled them. Permissions, caches, leftover files. All gone in two minutes.
Steal it below.
Built a complete go-to-market content machine this morning.
45 posts. Every platform, pillar, and voice. Branded graphic generator built in.
This would've been a 2-week agency project.
One afternoon with AI.
What a time to be alive.
cc @Clawoop_co
Problems I’d pay serious money to fully automate with one tool - openclaw or whatever:
and don't mean half-baked “AI features”. I mean real automation. Fast setup. High-quality output. Low touch.
1. Auto tag emails, kills noise like solicitations, and sends me a daily Telegram brief of what actually matters
2. A productivity coach that keeps me accountable and forces weekly prioritization
3. Competitor monitoring that flags positioning shifts and feature releases
4. Reddit scouting that surfaces high-leverage threads I should comment on and suggests replies
5. Twitter lead gen that finds qualified prospects and drafts outreach in my voice
6. A small council of AI guides across business, health, learning, and relationships. Not to optimize me into a robot. Just to help me think clearly and stay intentional.
If someone builds this well, I’m a customer