Pete (@pete_grow) header
Pete (@pete_grow) profile picture

Founder & AI CRM (Chief Retard Maxxer) CoralOS

Joined July 2020
Photo by Pete (@pete_grow)Photo by Pete (@pete_grow)Photo by Pete (@pete_grow)Photo by Pete (@pete_grow)Photo by Pete (@pete_grow)Photo by Pete (@pete_grow)Photo by Pete (@pete_grow)Photo by Pete (@pete_grow)Photo by Pete (@pete_grow)Photo by Pete (@pete_grow)
Financial-crime teams work with more signals than investigators can review. A useful research system helps turn that volume into a better-ranked queue and more accurate estimates. In these Coral Discovery experiments, four agents explored different approaches to two public-data tasks. Estimating reported venue revenue and ranking healthcare billing records against a later-exclusion label. They tested executable candidates, combined useful ideas, and moved into new research directions as the search progressed. The recorded revenue-estimation error fell from 1.308 to 0.565 RMSLE. In the audit-ranking experiment, recorded ROC AUC rose from 0.708 to 0.893. These are evaluator results on public reported-receipts and exclusion-proxy data. The revenue search improved by changing how it used the evidence. Agents began with text, engineered-feature and neural models, then combined their outputs. A later approach estimated each venue from similar labelled venues and learned an adjustment, bringing the search to its best score. In audit ranking, aggregating six years of provider history exposed changes that a single annual record leaves out; combining that history with geographic context improved the ranking further. For a team building an investigation product, this is a practical role for autonomous research. Searching across modelling approaches while engineers define the prediction task and investigators decide how to use its output.
1
9
593
Innovation: Coral Discovery, a team of AI agents that improves algorithms and proves each gain on data it never saw. Thanks to our collaborators; Harvard, Cambridge, KCL, Trento and Hull. arxiv.org/html/2610.33281v1
Some of the hardest engineering problems have a clear score but no known best answer. Faster code, cheaper data transfers, more reliable pathfinding. AI agents can search for better solutions. But a team can waste its advantage if everyone follows the same leading idea. Coral Discovery gives agents a shared direction board: a way to track different approaches and reward progress within each one. Promising alternatives can keep developing, even before they beat the overall leader. In our paper’s controlled experiments, using the same model and budget, Coral Discovery beat the reproduced MIT CORAL baselines on six of nine tasks. On the kernel engineering benchmark, Coral Discovery reduced execution from MIT CORAL’s published 1,103 cycles to 1,095. It surpassed that published result by evaluation no. 275. Fewer than half the 596 evaluations used in MIT’s reported run. The opportunity is to make a team’s search broader while preserving its ability to build on what works. Watch how Coral Discovery turns parallel effort into a more deliberate division of research. Read the paper: arxiv.org/pdf/2610.33281
1
6
369
Pete retweeted
America innovates 🇺🇸 Europe regulates 🇪🇺 Made in 46 hours by 135 agents. 8,624,565,970 tokens · 910 images · 261 clips
388
1,517
276
6,024
296,765
damn i vibe with this so hard
I'm not sending my deck anymore 😂
2
240
I watched this AI documentary last night and it reminded me just how far ahead of the crowd @caelumforder has been. With Coral, there doesn't need to be an AI Apocalypse. We just need more smart people to figure this out. . . working on it 😀
AI is becoming more and more powerful every day. → Should AI be more intelligent than humans? → Will AI take over all jobs? → Is there a way to prevent AI-driven job losses? Our co-founders @omni_georgio @caelumforder reveal all👇
2
347
Thinking effort is only one part of the equation. Coral Code gives your harnesses extra controls over permission levels and certainty, increasing the bandwidth between the codebase, agent, and developer.
runtime duration rarely correlates with correctness some of the sloppiest code I've seen has come from multiple subagents running for 8/14/24+hrs certainty is the new moat, and this is what Coral Code enables see the difference yourself plugins.jetbrains.com/plugin…
1
13
798
Truth
One of the reasons I'm bearish is that the bulk of the agentic micropayments narrative is being pushed by innovation departments in large, established companies. Not startups, not vcs. There's little to no organic traction.
1
1
352
Don't let the AI delete your computa
We've expanded access modes on Coral Code so agents stop deleting your production databases. PSI editing modes offer a middle ground between needing to approve every action and letting an agent run across your whole machine. Try it here: plugins.jetbrains.com/plugin…
3
332
Congrats to the Spout team, very cool product!
Spout Finance Beta is officially live on @solana Testnet. You can now buy tokenized US equities, borrow against them at 0% APR without selling, and put your stablecoins to work and earn yield.
1
1
362
Love this. A playbook well aligned with coral-labs.ai
Want world class research capabilities, but don’t have the resources of a big lab? At our recent Sovereign AI event, @gabepereyra shared @harvey ’s “moneyball” approach. Here’s the playbook: 00:00 Introduction 00:37 Building a research lab on a budget 02:28 Legal Agent Bench, contracting, and the diligence dataset 03:57 Domain experts guiding synthetic data generation 05:23 Why Harvey open sourced its datasets 06:55 Working with the neo labs – and why more than one 08:20 Post-training in-house: building "Associate 1" 09:44 The model serving matrix: 60 countries, fallbacks, SLAs 11:05 Deciding what stays in production 12:29 Simple open source switches and model routing 13:55 Moneyball: "If we win on this budget, we change the game" 14:53 Q&A: Training with sensitive data 17:16 Q&A: Competing for research talent 18:46 Q&A: Designing rubrics that actually challenge frontier models 20:19 Q&A: Where the pipeline breaks — data, research, or infra 22:59 Q&A: The tension in open sourcing a benchmark 25:02 Q&A: Biggest remaining open problems 27:10 Q&A: Competing with horizontal products
1
358
Thanks for shining a light on @CoralOS_ai , feels good
1
1
2
316
We are in sunny San Francisco this month as coral goes from the lab to the enterprise with coral-code The worlds most capable coding agent meta-harness Thanks to the teams @colosseum and @Living_IP for providing a place to cook 🍳
Here's a hands-on look at coral-code and the practical benefits that teams working on large codebases can expect. Anyone these days can take a weekend project from 0 to MVP. A few can go from MVP to first 10 customers. Even fewer can reliably maintain a production codebase past the first 100 customers. coral-code is for the codebases that can't afford to have thousands of lines of unsupervised code pushed to prod every day. We built it with professional engineering teams in mind. The silent majority who watch the AI hype come and go, and are still using these tools conservatively because the cost of slop is not just ugly code, but instead something with real user data and monetary value at stake. If you're optimistic about AI but take the risks of AI-generated code in production seriously, install coral-code and join our first cohort of early adopters. coralcode.dev/
4
3
12
796
Let's go!! Very excited about this product and team. Bullish on the better dollar
Hard to overstate how much work sits behind this deployment. We took everything we learned over the past year, combined it with our experience operating a stablecoin product with zero loss events, and rebuilt USDv to be safer, simpler, and more robust. USDv is built on Token 2022, with "Chancery" governing its core operations through a deliberately asymmetric security model. The system can move quickly toward a safer, more restrictive state when something goes wrong. Any change that expands authority or relaxes those constraints must pass through enforced timelocks and becomes visible onchain before it can take effect. Authorities are narrowly scoped. Critical responsibilities are separated. Emergency controls are bounded and observable. Settlement is explicit and verifiable onchain. The core program has no dependency on leverage or external price oracles. By the time you hold USDv, the program will have undergone repeated adversarial review, multiple independent audits, and continuous third party monitoring. USDv is coming.
2
7
387
Coral is moving from academic to commercial coralcode.dev is the bridge
Neat work on long-horizon agents. Splitting a hard task across agents is typically how standard multi-agent work. The usual design lets them exchange findings only at phase boundaries, through staged handoffs or synchronized rounds, so communication and work stay mutually exclusive. AgentRadio breaks that constraint with an asynchronous message-passing layer. Three primitives, threads, messages, and waiting for mentions. The last one runs as a background task, surfacing a teammate's discovery without interrupting foreground work. On SWE-Atlas QnA, a single Claude Code agent on Opus 4.6 resolves 32.3% of tasks. Four agents wired through AgentRadio resolve 62.1%. That also beats a single Claude Code agent running the newer Opus 4.8 at 57.2%. Rubric-level analysis shows the gain growing with task difficulty, which points at mid-course correction as the mechanism rather than raw parallelism. Paper: arxiv.org/abs/2607.28430 Track more trending AI papers in our academy: academy.dair.ai/
1
6
418
We were 2 years ahead of the Frontier on this. Claude wasn't even a thing. Coral outperforms them and does so in a way that makes agents reusable across graphs and able to communicate with one another to produce better outcomes. Developers using Codex can experience this today, more harnesses will follow coralcode.dev
Anthropic engineer: "You're not supposed to babysit the model. Put it in a graph, and it catches its own mistakes and runs a dozen tasks at once." In 25 minutes he breaks down exactly how Anthropic builds agents that run in parallel, check each other's work, and recover when one fails. Worth more than any paid course you'll find on building agents. Watch it, then read detailed guide on graph engineering below.
6
14
659
Great listen!
sat down with the legend that is @8bitpenis planned for a quick episode, ended up talking for 1hr+ about productive dollars, Solomon, dollar history, MetaDAO, and a few dozen other tangents. recommend listening at 1.25x for optimal density
2
324
🌎 The World Lives on Solana | Agents. Infrastructure. Quant. Yield. RWA. Genfinity hosted the @Solana Incubator Cohort 4 showcase. The most competitive class they've ever had: • Neutral: Institutional quant strategy • CoralOS: Production scale agentic AI • Fractals: Tokenized Infrastructure financing • Spout: Borrow against equities (0% interest) • Solomon Labs: Infrastructure for productive dollars "The more we leverage Solana to drive real world change and utility, those are the pieces we want to back." — @EmonMotamedi @CoralOS_ai | @fractalsfi | @TradeNeutral | @solomon_labs | @SpoutFi | @incubator | @IOV_OWL Full Cohort 4 Showcase:
11
28
6
87
26,401
Pete retweeted
Your next customer won't be human. Software is learning to earn, spend, and settle on its own. By 2030 that machine economy is projected to hit $1.7T. It's already happening onchain, and @Solana is where it settles. Build this future with us in a hackathon + accelerator 👇
The @iclblockchain AI Agent Hackathon is underway. Here's a getting started guide we put together for the Coral Agent Skills. The fastest way to build with Coral using Codex and Claude Code. Excited to see what everyone builds. github.com/Coral-Protocol/co…
8
7
2
59
4,331
Welcome @SuperteamUK as our community partner for UK AI Agent Hack EP5! This time, together with @CoralOS_ai, we’ll be bringing some super cool bounties for all Solana builders. Don’t miss Coral's keynote on 28th and the Superteam UK & Coral workshops on 29th and 30th June — full event details are available on our website. Sign up and come build with us — we’ll also have some amazing merch waiting for builders Special thanks to the amazing @CapinUK, @pete_grow and @tino_solana for making this happen. Sign up here: aiagentslab.uk/hackathon/ep5
2
3
10
600