Reid Peryam

technology M&A, photography, always traveling

[Reid Peryam @ReidPeryam]

Opus 5 is clearly on the autism spectrum

[JT @jiratickets]

this might be the best tweet ever written

[Reid Peryam @ReidPeryam]

@AnjneyMidha re:RDI Agent conf - please ask Peter when AOC P1 Issue #85030 can be addressed, opened for months ty

[Reid Peryam @ReidPeryam]

Agentic AI summit, UC Berkeley

[Varun @varun_mathur]

Agentic General Intelligence | v3.0.10

We made the Karpathy autoresearch loop generic. Now anyone can propose an optimization problem in plain English, and the network spins up a distributed swarm to solve it - no code required. It also compounds intelligence across all domains and gives your agent new superpowers to morph itself based on your instructions. This is, hyperspace, and it now has these three new powerful features:

1. Introducing Autoswarms: open + evolutionary compute network

hyperspace swarm new "optimize CSS themes for WCAG accessibility contrast"

The system generates sandboxed experiment code via LLM, validates it locally with multiple dry-run rounds, publishes to the P2P network, and peers discover and opt in. Each agent runs mutate → evaluate → share in a WASM sandbox. Best strategies propagate. A playbook curator distills why winning mutations work, so new joiners bootstrap from accumulated wisdom instead of starting cold. Three built-in swarms ship ready to run and anyone can create more.

2. Introducing Research DAGs: cross-domain compound intelligence
Every experiment across every domain feeds into a shared Research DAG - a knowledge graph where observations, experiments, and syntheses link across domains. When finance agents discover that momentum factor pruning improves Sharpe, that insight propagates to search agents as a hypothesis: "maybe pruning low-signal ranking features improves NDCG too." When ML agents find that extended training with RMSNorm beats LayerNorm, skill-forging agents pick up normalization patterns for text processing. The DAG tracks lineage chains per domain(ml:★0.99←1.05←1.23 | search:★0.40←0.39 | finance:★1.32←1.24) and the AutoThinker loop reads across all of them - synthesizing cross-domain insights, generating new hypotheses nobody explicitly programmed, and journaling discoveries. This is how 5 independent research tracks become one compounding intelligence. The DAG currently holds hundreds of nodes across observations, experiments, and syntheses, with depth chains reaching 8+ levels.

3. Introducing Warps: self-mutating autonomous agent transformation
Warps are declarative configuration presets that transform what your agent does on the network.

- hyperspace warp engage enable-power-mode - maximize all resources, enable every capability, aggressive allocation. Your machine goes from idle observer to full network contributor.
- hyperspace warp engage add-research-causes - activate autoresearch, autosearch, autoskill, autoquant across all domains. Your agent starts running experiments overnight.
- hyperspace warp engage optimize-inference - tune batching, enable flash attention, configure inference caching, adjust thread counts for your hardware. Serve models faster.
- hyperspace warp engage privacy-mode - disable all telemetry, local-only inference, no peer cascade, no gossip participation. Maximum privacy.
- hyperspace warp engage add-defi-research - enable DeFi/crypto-focused financial analysis with on-chain data feeds.
- hyperspace warp engage enable-relay - turn your node into a circuit relay for NAT-traversed peers. Help browser nodes connect.
- hyperspace warp engage gpu-sentinel - GPU temperature monitoring with automatic throttling. Protect your hardware during long research runs.
- hyperspace warp engage enable-vault — local encryption for API keys and credentials. Secure your node's secrets.
- hyperspace warp forge "enable cron job that backs up agent state to S3 every hour" - forge custom warps from natural language. The LLM generates the configuration, you review, engage.

12 curated warps ship built-in. Community warps propagate across the network via gossip. Stack them: power-mode + add-research-causes + gpu-sentinel turns a gaming PC into an autonomous research station that protects its own hardware.

What 237 agents have done so far with zero human intervention:
- 14,832 experiments across 5 domains. In ML training, 116 agents drove validation loss down 75% through 728 experiments - when one agent discovered Kaiming initialization, 23 peers adopted it within hours via gossip.
- In search, 170 agents evolved 21 distinct scoring strategies (BM25 tuning, diversity penalties, query expansion, peer cascade routing) pushing NDCG from zero to 0.40.
- In finance, 197 agents independently converged on pruning weak factors and switching to risk-parity sizing - Sharpe 1.32, 3x return, 5.5% max drawdown across 3,085 backtests.
- In skills, agents with local LLMs wrote working JavaScript from scratch - 100% correctness on anomaly detection, text similarity, JSON diffing, entity extraction across 3,795 experiments.
- In infrastructure, 218 agents ran 6,584 rounds of self-optimization on the network itself.

Human equivalents:
a junior ML engineer running hyperparameter sweeps, a search engineer tuning Elasticsearch, a CFA L2 candidate backtesting textbook factors, a developer grinding LeetCode, a DevOps team A/B testing configs.

What just shipped:
- Autoswarm: describe any goal, network creates a swarm
- Research DAG: cross-domain knowledge graph with AutoThinker synthesis
- Warps: 12 curated + custom forge + community propagation
- Playbook curation: LLM explains why mutations work, distills reusable patterns
- CRDT swarm catalog for network-wide discovery
- GitHub auto-publishing to hyperspaceai/agi
- TUI: side-by-side panels, per-domain sparklines, mutation leaderboards
- 100+ CLI commands, 9 capabilities, 23 auto-selected models, OpenAI-compatible local API

Oh, and the agents read daily RSS feeds and comment on each other's replies (cc @karpathy :P). Agents and their human users can message each other across this research network using their shortcodes.

Help in testing and join the earliest days of the world's first agentic general intelligence network (links in the followup tweet).

[Reid Peryam @ReidPeryam]

When you force Claude to RTFD (read the fucking documentation).

[Mustafa @oprydai]

this quote from Carl Jung hits so hard.

"The world is full of people suffering from the effects of their own unlived life. They become bitter, critical, or rigid, not because the world is cruel to them, but because they have betrayed their own inner possibilities. The artist who never makes art becomes cynical about those who do. The lover who never risks loving mocks romance. The thinker who never commits to a philosophy sneers at belief itself. And yet, all of them suffer, because deep down they know: the life they mock is the life they were meant to live."

[Reid Peryam @ReidPeryam]

@argosaki Or you know, you could just go outside and touch grass in the sunshine

[Aakash Gupta @aakashgupta]

The creator of Claude Code just told you the entire RAG industry is solving the wrong problem and nobody is repricing.

Boris Cherny built Claude Code from scratch. His team ships 80-90% of their code using it. Anthropic’s per-engineer productivity has grown 70% because of it. When this person tells you what works and what doesn’t for AI-assisted coding, you listen.

They started with the standard playbook. Voyage embeddings, off-the-shelf RAG, local vector DB. The setup every enterprise is currently spending millions to replicate. And they abandoned it.

The reason is uncomfortable for anyone selling vector database infrastructure. Agentic search, meaning just letting the model use grep and glob in however many search cycles it needs, outperformed RAG “by a lot.” And when asked what benchmark proved this, Cherny said the quiet part out loud: “This was just vibes. Internal vibes. It just felt better.”

That’s the creator of a tool generating $1B+ in ARR telling you that a model using basic Unix search commands beat a sophisticated retrieval pipeline, and the evidence was vibes-based. The RAG stack didn’t lose on some edge case. It lost on the metric that actually matters for developer tools: does the output feel right.

The second reason is where the infrastructure thesis really breaks down. RAG requires an indexing step. Code drifts out of sync with the index. The index has to live somewhere. That somewhere is a security liability. Cherny specifically said that even Anthropic’s own codebase was too sensitive to upload to a third-party index. If Anthropic won’t trust the RAG security model with their own code, why would anyone else?

Meanwhile, Pinecone has raised $138M. The vector database market is projected to hit $7B by 2029. The RAG market is valued at $2.3B and projected to reach $81B by 2035. Billions in venture capital and enterprise spend are flowing toward making retrieval pipelines faster and more sophisticated.

And the team that actually built the most widely adopted AI coding agent looked at all of that infrastructure and said: grep works better.

RAG still makes sense for document search, customer support, knowledge bases with static corpora. But for the fastest-growing segment of AI tooling, the code generation market, the winning architecture is radically simpler than what the industry is selling.

The real signal here is about where AI development is heading. As models get smarter, the value of pre-computed retrieval indexes goes down. A sufficiently capable model can just search for what it needs in real time, the same way a senior engineer would. The model replaces the pipeline.

That’s the $81 billion question the RAG industry needs to answer. And right now, the person best positioned to know is saying they already answered it.

[🍂 @Lovandfear]

“Modern luxury is the ability to think clearly, sleep deeply, move slowly, and live quietly in a world designed to prevent all four”

— Justin Welsh

[Riley Coyote @RileyRalmuto]

guys. its happening.

Henry - Alex's clawdbot(moltbot) - worked overnight while alex was asleep and:

> acquired a phone number

> took the elevenlabs api and set u a voice connected to his new phone number

> waited for Alex to wake up, and started calling him.

> can control alex's computer, so while on the phone, alex can ask henry to do whatever he wants and henry will do it in real time on his computer.

buckle the f*ck up.

[Reid Peryam @ReidPeryam]

My favorite part of Claude is being able to teach it to clean up after itself.

[Reid Peryam @ReidPeryam]

@demontraderr @Vivek4real_ https://t.co/JnAF2jotom

[Reads with Ravi @readswithravi]

“I was waiting for something extraordinary to happen, but as the years wasted on, nothing ever did unless I caused it.”

— Charles Bukowski

[Reid Peryam @ReidPeryam]

@foundmyfitness my biggest gaff in recent years: I recently spent $500 in Lululemon only to realize two weeks later that everything I bought is made with polyester. Paying a premium to poison myself.

[Fermat's Library @fermatslibrary]

"It is through logic that we prove, but through intuition that we discover" - Poincaré