Advay Gupta

Founder & Trader

1x exit
ex-Jane Street
Building across markets, AI and automation

[Advay Gupta @advaygupta]

interesting tweet from head of alignment to put out publicly -- right before IPO

[Evan Hubinger @EvanHub]

Jacob is correct here—we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.

[Advay Gupta @advaygupta]

@ArielKwiat I think Extra High is the best for anyone in yhe higher limit tiers

[Advay Gupta @advaygupta]

so hyped

[Advay Gupta @advaygupta]

the global IQ is now down 50%

[Advay Gupta @advaygupta]

I will die on this hill.

spending $700 a month on Anthropic ($200), OAI ($200), and xAI ($300) max plans is the HIGHEST RoI move for anyone in anything remotely adjacent to tech

[Douglas A. Boneparth @dougboneparth]

I’m taking a break from mental health so I can focus on the stock market.

[Advay Gupta @advaygupta]

through 210 this morning

[Advay Gupta @advaygupta]

yes semis = no ai for some reason

[Advay Gupta @advaygupta]

the SEC asking the banks who knew what & when after Citadel already

-> bought the ~$16B public book at a 10% haircut,
-> pushed $4B through nearly 100 block trades and
-> shed 80%+ of the risk:

[unusual_whales @unusual_whales]

BREAKING:

The SEC has launched an investigation into the trading of Leopold Aschenbrenner's Situational Awareness fund, per NYT.

The SEC sent subpoenas to multiple banks that gave Situational Awareness money to trade with, asking for details about the timing of trades and how Situational Awareness communicated with lenders.

[Advay Gupta @advaygupta]

yes semis = no ai for some reason

[Advay Gupta @advaygupta]

wasn't liquor

[Wall St Engine @wallstengine]

Ken Griffin’s Citadel has already shed more than 80% of the aggregate risk it took on from Leopold Aschenbrenner’s Situational Awareness portfolio.

Griffin on the exit:

“These included the largest intraday block trades of the year in 10 different names. In the United States, we quickly exited the vast majority of the vega risk, saving substantial capital in the process. Of the remaining positions in G10, only five have a gross market value greater than $250 million...”

[Jay Malavia @thejay]

Today, we celebrate one year of Kairos.

We pride ourselves on being forward-looking, but we want to take a moment to reflect on the year we’ve had and where we hope to go from here.

Kairos takes its name from the ancient Greek god of the opportune moment – the brief instant when an opportunity appears and must be seized.

He was depicted as a winged figure moving swiftly, with a single lock of hair falling over the front of his head and nothing at the back. As he approached, you had one chance to reach out and grasp that lock. But once he had flown past, there was nothing left to hold on to.

Kairos is the moment you have to reach for.

Ours is now.

The greatest opportunities require being early and having conviction. This principle is central to Kairos’s ethos, from the talent we hire, to the traders we empower, to our origins themselves.

Kairos began as a belief – a belief that prediction markets were becoming real financial markets, and that the people trading them would eventually demand technology built to the same standard.

So we spent the past year building for those traders. That means building technology that is faster, more opinionated, more curated, and designed for people who want to win. It means ensuring that our traders have the information and tools to act the moment an opportunity appears.

Most of the industry is building applications that make event contracts feel like another form of gambling.
We don’t think prediction markets should feel like slot machines. We think a product should make its traders better over time. Traders hand Kairos their decisions and trust. We’ve treated that as our first obligation since the beginning, and we will for the life of the company.

We see a future in which prediction markets develop into a genuine asset class used by serious traders and institutions to price information, express views, and manage risk.

To get there, the market will need better data, better pricing, better execution, and technology built specifically for how these instruments trade. We intend to supply it.

One year ago, Kairos was little more than a few words scribbled on a Cboe notepad. Since launching our open beta two months ago, we’ve processed millions of dollars in trading volume and are growing exponentially.

We got a hand on the lock of hair.

[Advay Gupta @advaygupta]

@justmazer @mylespatel

[The Kobeissi Letter @KobeissiLetter]

BREAKING: Reddit stock, $RDDT, surges over +11% on news that the company is being added to the S&P 500.

[Advay Gupta @advaygupta]

@Reddit

[Advay Gupta @advaygupta]

@JustinGorya wow! what was the prompt you used?

[Di🇻🇳 @diluois93]

0DTE options now account for more than half of S&P 500 index options volume on many trading days, per Cboe data

[Advay Gupta @advaygupta]

Credit where due: the @xai engineers at Grokathon confirmed the root cause on the spot and said a fix is coming. That kind of access to the team is exactly what made the event worth it.

The network that found the bug is open source: https://t.co/CJzkstQRBR

It is still the only TV station we know of that files its own bug reports.

[Advay Gupta @advaygupta]

BUG REPORT: We built Grok Box at Grokathon, an autonomous AI news network that researches, writes, renders, and broadcasts itself live on X. Around hour three it surfaced a production bug in the xAI API that was structural.

Here is what we found.

[Advay Gupta @advaygupta]

The root cause, confirmed by xAI engineers at the event: x_search and web_search run a server side agent loop, and tool_choice "required" applies to every iteration of that loop. The model is never allowed a final turn to emit its answer. The tool budget runs out and the request dies as an opaque 500 instead of failing gracefully.

Workaround: tool_choice "auto" with explicit search instructions in the prompt. If you force tool choice with the agentic search tools and see random 500s, this is why.