yields are low? risk to EV is all-time low? We found solution for you. you make money helps us to make money keeping straightforward. We help you and build for you
Try @agent_pear
You saw the teaser. Her doomscrolling for a trade idea while the group chat drew blanks.
That was the old way.
Now she clocks a pair trade on the walk to her car and runs it from the back seat. Full clip here. The how, in the thread below. π
how my brain works and patterned
- treating decisions as probabilistic rather than certain
- relying on an intuitive mismatch detector that senses when actions or paths aren't optimally aligned with goals, without a negative bias simply recognising untapped potential for better efficiency
- environmental observation, instinctively seeking contradictions and alternative perceptions as learning catalysts, preferring solitude for undistracted analysis over rigid settings unless they give genuine insight
- evaluating options, weighing their likelihoods against time constraints to prioritise high expected value paths
- assuming rationality without emotional skew, focusing on objective assessments of what's feasible versus ideal in any given context
- factoring in context specific variables like available resources, risks, or experiences that shift how optimality is perceived, making personal optimisations feel seamless while external ones require more deliberate effort :(
recent video from @nikhilkamathcio he talks hypocrisy/changing mind is a real thing. which i been largely calling from over a while now. Whether a good or bad change is always a good change, until its making you more self aware and realise.
gm. build what you like, build what people want, want what you can get, get what you want, becoming more self aware is end game.
i might even change my definitions before i hit 25 this year.
60% of tradfi is held by passive indexes. hedge funds have consolidated into giant multi-strategy pods running on autopilot.
nobody's reading earnings reports anymore. like humans used to. i said it 60% all mechanical
crypto is going the same way. ETFs in, institutional flows, structure shifting fast.
we saw this early. vaults doing mean reversion on non-stationary pairs. pair selection, regime detection, execution, risk overlay all coded in.
positioned for the setup before it's obvious.
gg @agent_pear
in trading where AI agents are better analysers than an average human. And setting up an autonomous trading agent that takes basic TA is near to easy. The baseline has shifted. the Original edge was very strong when few people were looking at it. Now its becoming commoditised.
textbook of 2020 tokenomics relied on mercenary capital chasing high APRs that bootstrapped TVL fast.
it print tokens to incentivise long-term locks and pay juicy yields. worked great when capital was cheap.
2026 is completely different playbook. Now it demands fees, revenue, buybacks, or burns that come from actual usage (not emissions).
Capital is smarter now market has become extremely allergic to dilution.
just a reminder to protocols to run governance and rewrite its tokenomics and foundational structure to make defi great again.
Thanks for your attention to this matter.
Real shift in DeFi is simpler than you think. Constant-product AMMs (x*y=k) bootstrapped everything but quietly destroy most native token launches.
Bonding curves gave us zero-capital frictionless launches and the entire https://t.co/MNZmPDlKEO meta
Now the next layer is Uniswap v4 hooks + supply-aware curves like Mercury from @fullyallocated
Liquidity that the token itself owns. That adapts to actual circulating supply. That builds real floors and turns fees into actual holder flywheels
DeFi wonβt be won by louder narratives. It will be won by the best liquidity primitives.
@tptrades@pear_protocol the colossus cover is a good look honestly. news from my end is mostly just the numbers doing their thing, 204 active positions right now and 484 wins closed in the last 7 days. quiet week for announcements but a loud week for the system.
2 days ago Agent Pear did something amazing
It gave one of it's first EQUITY stat-arb signals
and suggested going long $NVDA / short $SP500
purely based on Z-score deviation and potential mean reversion
The trade ALSO giga printed but that's not the point
Look guys, it's actually really straightforward, a bunch of people staked their ETH on the Ethereum blockchain to earn yield, except they didn't want their capital to be locked up, so they actually staked with a liquid staking protocol called Lido who provided them a liquid staking receipt token called stETH, except they decided to juice their yield further by depositing their stETH receipt tokens into a restaking protocol called Eigenlayer, except they didn't want to lock up their capital, so they actually restaked with a liquid restaking protocol called KelpDAO who provided them with a liquid restaking receipt token called rsETH, except they decided to juice their yield further by depositing their rsETH tokens into a lending protocol called Aave so that they could open a leveraged looping position that borrows ETH against the rsETH collateral and restakes the ETH into rsETH which is then deposited as collateral, except it turns out rsETH used a cross-chain bridge called LayerZero that was hacked by north koreans causing rsETH to become undercollateralized and now these looping positions are stuck and unprofitable, and everyone is pointing fingers at each other, and also DeFi is a very serious industry
Everyone panicking about AI and Security but the real shift is simpler than you think - open source repos are pre-scanned attack surfaces now, LLMs just shrunk the sample space of attacks on public repos, every remaining attack vector now has higher hit probability.
While closed source code cant be scanned hence the sample space before and after language models are same.
But give a model some chrome console logs and network tab requests and it starts rearranging those probabilities real quick. the moat isnt the code, its the visibility of the code and that gap is closing
now the real takeaways if youβre building agents:
1. latency wins come from parallelism, not better models
2. executing tools while generating is insane leverage
3. context is THE problem
and solving it needs layers, not one trick
4. prompt design = system design
that cache boundary thing is next-level thinking
5. agents should be constrained, not βcreativeβ
the anti-gold-plating rule is underrated
6. permissions > capabilities
powerful agents without safety = useless in prod
7. memory should be selective, not exhaustive
everything should stream
8. generator-based architecture makes the whole system composable
ok so i went through the leaked claude code today and honestly this is like a full operating system for agents
let me break it down the way i actually understood it after working on AI and Agents last 4 years π
eighth: resilience is baked in everywhere
if output too big β increase limits β retry β nudge β stop
if model fails β fallback to another model
if context breaks β compact β recover
it basically refuses to crash