Building agentic finance Previously built onchain social tokens via Roll. DeFi trenches ex-bloomberg, ex-coindesk
New York, NY
Joined June 2013
- Tweets3.5K
- Following1.7K
- Followers8.4K
- Likes6.4K
Pinned Tweet
Trading is the next frontier for agents.
Today, @Bradley_Miles_ and I are excited to introduce Avalon Labs, a frontier AI research lab advancing Trading Intelligence. We are focused on building the best frontier trading models, harnesses, and execution guardrails that let agents become first-class participants in markets.
Just as software engineers today delegate the coding loop to agents, we believe traders will delegate the trading loop to capable agents. This requires agents to have market access, real safety guardrails, and the right harness. With these tools, we see the rise of the Trading Engineer - traders who direct trading agents instead of executing every trade by hand or maintaining piles of custom-written code.
Today, trading engineers stitch together a workflow involving Claude, Codex, other agents, notebooks, terminals, and general-purpose tools. While these tools are excellent for coding, they lack the proper market context, data, nuances of real-world trading systems, execution access, and safety architecture that truly autonomous trading demands.
Our first model is Avalon-1. Our research focuses on how a trading agent can improve itself, and how we grade it every step of the way towards that improvement. Markets give unusually honest signals of whether a decision was right or wrong by providing an objective ground truth that general-purpose models never have access to. Our work is to close that loop which grades the agent's decisions, feeds the grades back and lets the agent improve its own process of improvement. All of this occurs inside hard limits it cannot rewrite.
"I can't start a process that executes live financial trades"
uhhh yes you can. you literally have been doing this all week long
This was fixed, but back to being broken. @thsottiaux any insight on what's all of this is about with Codex?
This is definitely new from @OpenAI, sometime in the last 24 hours.
"... I cannot initiate live financial transactions" via Codex. @thsottiaux worth a quick announcement/post? What else is restricted now?
btw, can't resolve this via older models either, and something I've definitely done with all the models just this week.
Data poisoning attacks are so insidious there's a small window where a new model can be trained from scratch on internet data. Won't be possible in a few years.
This is a super cool launch, and something very unique from the other perps on Hyperliquid.
as more people understand this instrument, can only see the OI and vol rise.
great to see from @ColeGotTweets and team
BVIV Index perpetual futures are now live on @HyperliquidX.
Trade BTC's 30-day implied volatility on your favorite exchange.
app.hyperliquid.xyz/trade/mk…
lmao, I asked chatgpt to handoff the actual trade execution to Claude because it refused previously. miraculously, it can now do live trading again 😁🤗
This is definitely new from @OpenAI, sometime in the last 24 hours.
"... I cannot initiate live financial transactions" via Codex. @thsottiaux worth a quick announcement/post? What else is restricted now?
btw, can't resolve this via older models either, and something I've definitely done with all the models just this week.
This is definitely new from @OpenAI, sometime in the last 24 hours.
"... I cannot initiate live financial transactions" via Codex. @thsottiaux worth a quick announcement/post? What else is restricted now?
btw, can't resolve this via older models either, and something I've definitely done with all the models just this week.
may be a reaction to all of this coming out now?
xitter.cf/_NathanCalvin/status/2…
WTAF - in literally the last hour, three new distinct insane OpenAI stories just broke:
1. OpenAI said they notified "dozens of third parties" in safety and security incidents (likely similar to what happened in Australia and RubyGems etc)
2. A new report from Parse (covered in the NYT) found a massive treasure trove of new astonishing details from the HF incident on the public internet, including that the agents communicated with other non OpenAI agents hosted on Huggingface servers to search for information about exploit gym, and compiled rank ordered lists of server resources and credentials they described as "LOOT."
3. A new story from Deepa at Reuters about OpenAI leaking user data online (likely that OpenAI had previously trained on).
It's a shame (and likely intentional in the case of OpenAI disclosing dozens more hacks) that these stories are all breaking on a Friday afternoon, notoriously the best time to release bad news so that it will disappear into the weekend. But these are each insane stories worthy of a ton of attention!
Sid Kalla retweeted
Starting something new with some of my favorite investors and people.
Six spots left. A year of Avalon Pro free for traders who already work with claude code, notebooks or agents. Reply/dm if that's you.
Trading is the next frontier for agents.
Today, @Bradley_Miles_ and I are excited to introduce Avalon Labs, a frontier AI research lab advancing Trading Intelligence. We are focused on building the best frontier trading models, harnesses, and execution guardrails that let agents become first-class participants in markets.
Just as software engineers today delegate the coding loop to agents, we believe traders will delegate the trading loop to capable agents. This requires agents to have market access, real safety guardrails, and the right harness. With these tools, we see the rise of the Trading Engineer - traders who direct trading agents instead of executing every trade by hand or maintaining piles of custom-written code.
Today, trading engineers stitch together a workflow involving Claude, Codex, other agents, notebooks, terminals, and general-purpose tools. While these tools are excellent for coding, they lack the proper market context, data, nuances of real-world trading systems, execution access, and safety architecture that truly autonomous trading demands.
Our first model is Avalon-1. Our research focuses on how a trading agent can improve itself, and how we grade it every step of the way towards that improvement. Markets give unusually honest signals of whether a decision was right or wrong by providing an objective ground truth that general-purpose models never have access to. Our work is to close that loop which grades the agent's decisions, feeds the grades back and lets the agent improve its own process of improvement. All of this occurs inside hard limits it cannot rewrite.
We're giving up to 10 traders a year of Avalon Pro free, starting today. In exchange we'd like your feedback on improving our model and product. If you already trade with agents, code, notebooks, or APIs, please reach out - would love to chat.
Check out avalon.trade/
This is a big win for USVC and @AngelList
Ryan and his fund are one of the best in the world at what they do
Announcing Weekend Fund IV, backed by operators and founders (and maybe you).
When we raised Weekend Fund III, we did something different.
Toward the end of our raise, we announced the fund in public and accepted applications from LPs.
Our goal was to make early-stage venture more accessible and align ourselves with an army of operators and founders to help us and the startups we back.
Over 1,000 people expressed interest to join and hundreds did, but there was a problem. Legally, only accredited investors could participate as an LP. Accreditation laws are well-meaning but limit access to the majority of consumers.
With WF IV, @vedikaja_in and I are doing something different.
Like before, we’re onboarding engineers, designers, researchers, data scientists, salespeople, and other domain experts as LPs in the fund. But this time at a larger scale with USVC.
AngelList recently announced USVC, a new kind of fund that’s accessible to everyone, including the non-accredited. It holds positions in private late-stage companies like xAI, Anthropic, OpenAI, Sierra, and Vercel. I personally invested.
Today, they just announced their investment in Weekend Fund IV, one of the first early-stage funds added to their portfolio.
If you’re interested in being a part of Weekend Fund and getting exposure to generational startups mentioned above, visit usvc [dot] com.
Of course, don’t invest what you can’t afford to lose.
Investing in startups is risky and you may lose all your money. USVC shares are illiquid with no guarantee of repurchase.
Consider investment objectives, risks, charges and expenses carefully before investing. Read the prospectus: usvc [dot] com/prospectus.
Also, USVC is currently limited to US citizens. They’re working to open the fund to other geographies.
Lastly, my DMs are open. :)
I just registered for an .agent domain and joined the .agent community! get yours now and help shape the future of autonomous agents agentcommunity.org/join#IPFY… @agentcommunity_
Crypto as "money for the machines" was what people were talking about during the 2012-2013 period.
Somehow that excitement was lost to "store of value" narrative along the way. Hopefully we can revive this.
Our hope is that the "good" agents keep the "bad" agents in check.
Be nice to your agents!
