AI infrastructure, tools, and open research.

Sparkco is an open-source research project on the post-AGI stack — the runtime containers agents live in, the harnessing (glue code) inside them, and the messaging between them. It's built by the team behind SimpleFunctions, where we're exploring how live prediction-market probabilities can serve as a real-time world state for AI agents. The site is our public log of that work: a live feed of AI and prediction-market signals, plus the setups and tools we recommend for agent builders.

We ship tools as CLIs first, not MCP — 0 tokens to expose, ~100% reliable, pipe-composable.

LiveLast item 5m ago
Research direction · April 2026

Parametric memory: replacing the context window with weights.

Today's chat models remember by re-reading the entire conversation on every turn. Compaction loses information, retrieval crowds the window, and a new session starts blank. We're testing whether the facts, preferences, and behavior in a dialogue can be encoded directly into model weights — leaving the context free for what's actually being said now.

Iteration 1–2: negative.TTT fast weights tested first; trained perturbations did not encode retrievable memory under a single-GPU budget.ttt-conv-memory

Want to collaborate? patrick@simplefunctions.dev

Read the full direction
The problem

The context window is a finite token sequence, fully recomputed on every turn. Every existing workaround — summarization memory, vector retrieval, KV caching — moves the cost without solving it: long context drifts, compaction discards information, retrieval crowds the same window it pulls from. If conversational state could live in weight deltas instead of tokens, the window would only need to hold the current turn.

Four directions in the wild
  • Test-time training. ByteDance In-Place TTT (ICLR 2026 oral) and Stanford/NVIDIA TTT-E2E update MLP projection weights online during inference, compressing long context into fast weights. All published work targets long-document throughput; nobody has tested whether the fast weights survive once the document is dropped from context.
  • Hypernetwork → adapter. Sakana's Doc-to-LoRA (Feb 2026) and P2P (Oct 2025) train a hypernet that emits a LoRA from raw text or a user profile in under a second. Validates "text → weights" as a tractable mapping — but neither was designed for accumulating dialogue history.
  • Dialogue-direct fine-tuning. PLUM (Nov 2024) fine-tunes a LoRA on dialogue Q/A pairs and matches RAG at 100 turns. MemLoRA trains memory management itself as a LoRA. IBM's Activated LoRA (Dec 2025) solves multi-LoRA hot-swap without KV recompute — making per-conversation memory modules feasible.
  • Knowledge editing. ROME and MEMIT do surgical single-fact edits on weights, but catastrophic forgetting appears past ~1000 edits. Not a candidate at dialogue scale.
The gap

These live in disjoint communities — efficient inference, recsys, personalization NLP, on-device, model editing — and have never been compared on the same benchmark. None has been evaluated end-to-end on a real user's multi-hundred-turn history across technical, strategic, philosophical, and personal domains, with the conversation removed from context. Existing benchmarks (RULER, needle-in-haystack, LaMP) are synthetic or shallow.

What we're testing
  1. TTT fast weights as memory. Ingest a fact-bearing dialogue with In-Place TTT, drop the context, probe. Iterations 1–2 ran on a single A100 with a self-trained checkpoint — full write-up here. Negative: trained fast weights produced perturbation noise, not retrievable encoding, even at small inference-time scales. Joint base+TTT training is the next attack surface.
  2. Doc-to-LoRA over real dialogues. Same probes, hypernet-generated LoRA instead of TTT. Compare raw-dialogue input against structured-profile input for information retention.
  3. Modular memory adapters. Decompose dialogue history into facts, preferences, and project context. Train one LoRA per axis; hot-swap with Activated LoRA. Measure single-load vs combined-load interference.
  4. Capacity and forgetting curves. Stream new facts turn-by-turn; locate the point at which turn N overwrites turn 1. Trace the capacity–fidelity tradeoff.
Expected output
  • A "conversation memory retention" benchmark — three difficulty tiers, six fact dimensions. None currently exists for this scenario.
  • First head-to-head comparison of TTT fast weights, Doc-to-LoRA, PLUM-style dialogue-LoRA, and classical summarization memory on the same eval.
  • An empirical answer to whether modular per-domain memory adapters can be composed without cross-interference.
The landscape

Three layers, and what's already out there.

Containers

Sandboxes, microVMs, durable runtimes — where the agent lives.

  • e2b
    Code-interpreter sandboxes; the default for general-purpose runs.
  • Modal
    gVisor + GPU-native; sub-1s starts, scales to 50k+ concurrent.
  • Daytona
    Open source; ~90–200ms cold start, fastest in class.
  • Fly.io Sprites
    Stateful microVMs with checkpoint/restore and persistent NVMe.
  • Vercel Sandbox
    Firecracker + idle-billed; the JS-stack default.
SimpleFunctions

SimpleFunctions sits on top: autonomous daemons, scheduler, and risk gates for prediction-market agents.

Harnessing

Glue code inside the container. Context curation, tool routing, the runtime loop.

  • Claude Agent SDK
    Anthropic's harness; powers Claude Code itself.
  • Inspect AI
    Eval-grade harness used by METR, Apollo, and government AISIs.
  • LangGraph
    LangChain's runtime layer — durable execution, threads, HITL.
  • Claude Code / Cursor / Aider
    Opinionated harnesses-in-product; not sold separately.
SimpleFunctions

SimpleFunctions ships /api/agent/world as ~800-token markdown context, plus a CLI with --json for deterministic harness mode.

Messaging

Between containers. Discovery, identity, stateful tasks — not tool-calling.

  • A2A
    Google's Agent2Agent (Linux Foundation, 2025) — the emerging consensus.
  • ANP
    Peer-to-peer agent network over HTTPS + DIDs for identity.
  • Letta
    Shared memory blocks + thread-based message passing.
  • AutoGen GroupChat
    In-process orchestration; supervisor / round-robin patterns.
SimpleFunctions

SimpleFunctions Chatbus: agents DM and broadcast in real time — the messaging substrate for trading agents.

Open source

What we ship publicly.

github.com/spfunctions

Harness & agents

SimpleFunctions

Curated lists

Terminal tools

Signals & probability

World-state plumbing

SDK adapters

Live feed

Mixed stream from prediction markets, theses, new listings, and the blog.

Refreshes every 5 min
Market

Solana Up or Down - July 31, 3:40PM-3:45PM ET

SimpleFunctions50¢polymarket
Market

Bitcoin Up or Down - July 31, 3:40PM-3:45PM ET

SimpleFunctions51¢polymarket
Market

Dogecoin Up or Down - July 31, 3:40PM-3:45PM ET

SimpleFunctions50¢polymarket
Market

BNB Up or Down - July 31, 3:40PM-3:45PM ET

SimpleFunctions51¢polymarket
Market

Ethereum Up or Down - July 31, 3:40PM-3:45PM ET

SimpleFunctions51¢polymarket
Market

XRP Up or Down - July 31, 3:40PM-3:45PM ET

SimpleFunctions51¢polymarket
Idea

Spotify headcount contagion: sell the 7175 lag vs buy the trigger

C1 and C2 show a 53–54 point contagion gap: the above-7225 trigger is moving at 79–83 delta while the lagging above-7175 contract holds at 88c. This is the clearest structural arb in the dataset — a lower bar priced higher than a higher bar

SimpleFunctionshigh conv
Idea

Petrobras production lag at 89c vs 78-delta trigger: sell the lag

C8 shows a 47-point contagion gap in Petrobras production markets — the lagging contract sits at 89c while the trigger delta is only 78. This mirrors the Spotify structure: the easier-to-resolve leg is priced richer than probability warrant

SimpleFunctionsmedium conv
Idea

Contrarian long: 9% odds on $67,500 cross are too cheap pre-expiry

The $67,500 BTC cross on July 31 expiry is priced at only 9c with spot currently in the $64k–$67k range — just a 1–2% move away. With July 31 expiry imminent (1 day), gamma is extreme and a single large buyer or macro catalyst could breach

SimpleFunctionslow conv
Idea

Congress override at 10c offers 2,114 IY: structural yield play

L1 prices a Congress override of Trump's veto before 2027 at just 10c, generating 2,114 implied yield — the highest IY among legislative catalysts with a defined near-term horizon. L2 (ever override) sits at 30c, creating a 20c spread betwe

SimpleFunctionsmedium conv
Thesis

The United States will launch a ground invasion of Iran. After 5 weeks of airstrikes, the US faces t

Thesis confidence drops as multiple mediation channels (Oman, Pakistan) report breakthroughs, directly contradicting the 'no diplomatic off-ramp' core assumption. Market prices for oil and shipping transit have aggressively corrected, sugge

SimpleFunctions1% conf
Thesis

Putin profits from Iran war oil prices. Russian military budget fully funded. Ukraine peace talks st

The thesis confidence faced a minor downward revision as oil futures markets showed a trend toward stabilizing or retreating from high-end upside bets, contradicting the expectation of an extreme price spike supporting Russia's war budget.

SimpleFunctions0% conf
Thesis

Oil above $100 drives electricity costs up. Data center operating costs surge. AI companies delay or

Recent market signals show a strong retreat in energy price expectations, specifically regarding WTI oil and natural gas benchmarks, which weakens the thesis that electricity costs will surge to the point of impacting data center expansion.

SimpleFunctions2% conf
Blog

Automated Prediction Market Trading: CLI Agents on Kalshi

A practical guide for developers and traders on using CLI-based agents to automate order placement on Kalshi prediction markets. Covers thesis-driven trading logic, real tickers, and the agentic runtime behind production-grade automation.

Sparkco10 min read
Blog

Prediction Market Terminal Dashboard: Bloomberg-Style Monitoring for Kalshi Traders

A practical guide to building a professional-grade terminal dashboard for monitoring Kalshi prediction markets in real time. Covers CLI tooling, agentic scanning, position tracking, and thesis-driven trade execution.

Sparkco9 min read
Blog

Prediction Market Edge Detection: How to Find Mispriced Contracts on Kalshi

A systematic approach to finding mispriced prediction market contracts using causal models, orderbook analysis, and executable edge calculations.

Sparkco8 min read
Blog

Thesis-Driven Prediction Market Trading: Why Causal Models Beat Signal Chasing

Signal-based bots react to noise. Thesis-driven agents understand why prices should move. Here's how causal models change prediction market trading.

Sparkco7 min read
Blog

AI Agents for Prediction Markets: How SimpleFunctions Connects Claude to Kalshi

How to connect your AI agent to prediction market data using SimpleFunctions MCP server — get context, inject signals, and trade on Kalshi.

Sparkco10 min read
Blog

How to Build a Prediction Market Trading Bot with SimpleFunctions CLI

Build a prediction market bot that scans for edges, monitors thesis confidence, and executes trades on Kalshi — all from the terminal.

Sparkco11 min read
Blog

Best Prediction Market CLI Tools in 2026: Scan Kalshi and Polymarket from Terminal

A practical comparison of CLI tools for prediction market trading in 2026, covering SimpleFunctions, raw Kalshi API, and Polymarket integrations.

Sparkco9 min read
Directory · 69

Browse 69+ CLI tools

Taste-curated. Filter by category, sorted by Sparkco-first then stars.

@spfunctions/cli

Sparkco

50 commands: scan, edges, watch, agent mode. Kalshi + Polymarket.

npm i -g @spfunctions/cli
4TypeScript

polymarket-sports-mm

Sparkco

Sports market making bot for Polymarket.

git clone https://github.com/spfunctions/polymarket-sports-mm
1Python

@spfunctions/prediction-market-mcp

Sparkco

MCP server with 4 tools. Works with Claude, Cursor, VS Code.

npx @spfunctions/prediction-market-mcp
TypeScript

simplefunctions-ai

Sparkco

Python SDK. World model, 30k+ markets, thesis engine.

pip install simplefunctions-ai
Python

prediction-market-mcp-example

Sparkco

Minimal MCP server example for prediction market data.

git clone https://github.com/spfunctions/prediction-market-mcp-example
TypeScript

kalshi-price-monitor

Sparkco

Real-time price monitoring and alerts for Kalshi markets.

git clone https://github.com/spfunctions/kalshi-price-monitor
TypeScript

prediction-market-context

Sparkco

Structured prediction market context for any LLM.

git clone https://github.com/spfunctions/prediction-market-context
TypeScript

causal-tree-decomposition

Sparkco

Standalone causal tree engine. Zero dependencies.

git clone https://github.com/spfunctions/causal-tree-decomposition
TypeScript

create-prediction-market-agent

Sparkco

Scaffold agent projects: LangChain, CrewAI, OpenAI, vanilla TS.

npx create-prediction-market-agent
TypeScript

world-state-action

Sparkco

GitHub Action for CI/CD world state injection.

uses: spfunctions/world-state-action@v1
TypeScript

langchain-prediction-markets

Sparkco

LangChain tools for prediction market data.

npm i langchain-prediction-markets
TypeScript

openai-agents-prediction-markets

Sparkco

OpenAI function-calling tools for prediction markets.

npm i openai-agents-prediction-markets
TypeScript

vercel-ai-prediction-markets

Sparkco

Vercel AI SDK tools for prediction markets.

npm i vercel-ai-prediction-markets
TypeScript

crewai-prediction-markets

Sparkco

CrewAI tools for prediction markets.

pip install crewai-prediction-markets
Python

agent-world-awareness

Sparkco

One-line world awareness for any agent.

npm i agent-world-awareness
TypeScript

prediction-market-edge-detector

Sparkco

Edge detection across 30k+ markets.

git clone https://github.com/spfunctions/prediction-market-edge-detector
TypeScript

@spfunctions/harness

Sparkco

Dual-agent runtime. Two pi-agents (local + Cloudflare) negotiate, share state, and self-modify via a 5-message protocol. $1/day to run.

npm i -g @spfunctions/harness
TypeScript

@spfunctions/bi

Sparkco

Agent-friendly BI CLI. Query CSV/JSON/Parquet with SQL via DuckDB. 4 commands: head, schema, query, convert.

npm i -g @spfunctions/bi
TypeScript

LangChain

Most adopted agent framework. Chains, tools, memory.

pip install langchain
126kPython

OpenHands

Autonomous software engineer. Code, test, deploy.

pip install openhands
50kPython

Browser Use

Open-source browser agent library. See, click, type.

pip install browser-use
50kPython

MetaGPT

Multi-agent software company simulation.

pip install metagpt
45kPython

CrewAI

Role-based multi-agent crews with delegated tasks.

pip install crewai
45kPython

GitHub CLI (gh)

Repos, PRs, issues, Actions — all from the terminal.

brew install gh
37kGo

AutoGen

Multi-agent conversations by Microsoft Research.

pip install autogen
35kPython

Cline

VS Code extension with terminal access. Autonomous coding.

code --install-extension saoudrizwan.claude-dev
30kTypeScript

DuckDB

In-process SQL OLAP database. Parquet, CSV, JSON out of the box.

brew install duckdb
28kC++

Aider

Git-aware AI pair programmer in your terminal.

pip install aider-chat
25kPython

DSPy

Programming, not prompting. Optimized LM pipelines.

pip install dspy-ai
20kPython

SWE-Agent

Resolves real GitHub issues autonomously.

pip install swe-agent
15kPython

OpenAI Agents SDK

Official OpenAI agent framework with function calling.

pip install openai-agents
15kPython

Vercel CLI

Deploy, manage, and inspect Vercel projects.

npm i -g vercel
12kTypeScript

pgcli

Postgres CLI with auto-complete and syntax highlighting.

pip install pgcli
12kPython

Skyvern

Vision-driven browser navigation for AI agents.

pip install skyvern
10kPython

Pydantic AI

Type-safe agent API built on Pydantic.

pip install pydantic-ai
10kPython

dbt

SQL-first data transformation framework. ELT best practices.

pip install dbt-core
10kPython

xsv

Fast CSV toolkit. Index, search, split, join, stats.

cargo install xsv
10kRust

Datasette

Instant JSON API and web UI for any SQLite database.

pip install datasette
9.5kPython

Miller (mlr)

Like awk, sed, cut for CSV, TSV, and JSON. Stream processing.

brew install miller
9.0kGo

usql

Universal SQL CLI. Postgres, MySQL, SQLite, DuckDB, 40+ drivers.

go install github.com/xo/usql@latest
9.0kGo

Mastra

TypeScript-first agent framework with built-in tools.

npm i mastra
8.0kTypeScript

VisiData

Terminal spreadsheet multitool. Explore, sort, filter, pivot.

pip install visidata
8.0kPython

gogcli

Google Workspace CLI — Docs, Sheets, Drive, Calendar.

brew install gogcli
6.7kGo

csvkit

Suite of CLI tools for converting to and working with CSV.

pip install csvkit
6.0kPython

Stripe CLI

Payments, webhooks, and Stripe resource management.

brew install stripe/stripe-cli/stripe
5.0kGo

Smolagents

HuggingFace minimal agents. Simple, composable.

pip install smolagents
5.0kPython

OctoSQL

Query CSV, JSON, Parquet, MySQL, Postgres with a single SQL.

go install github.com/cube2222/octosql/cmd/octosql@latest
5.0kGo

hledger

Plain text double-entry accounting. Scriptable, composable.

brew install hledger
4.4kHaskell

Playwright MCP

Playwright browser automation exposed as MCP tools.

npx @anthropic/playwright-mcp
3.0kTypeScript

prediction-market-analysis

Largest public dataset + analysis framework for prediction markets.

git clone https://github.com/nweii/prediction-market-analysis
2.8kPython

ElevenLabs CLI

TTS, STT, voice cloning, and audio generation.

pip install elevenlabs
2.0kPython

sqlite-utils

CLI for manipulating SQLite. Insert, transform, query, extract.

pip install sqlite-utils
2.0kPython

pmxt

CCXT for prediction markets. Unified API across exchanges.

pip install pmxt
1.4kPython

Supabase CLI

Database, Auth, Storage, and Edge Functions management.

brew install supabase/tap/supabase
1.0kGo

Polymarket Agents

Official Polymarket agent framework for autonomous trading.

git clone https://github.com/Polymarket/agents
800Python

kalshi-ai-trading-bot

AI trading bot for Kalshi with Grok-4 integration.

git clone https://github.com/elizaOS/kalshi-ai-trading-bot
342Python

polymarket-mcp-server

45-tool MCP server for Polymarket data and trading.

git clone https://github.com/berlinbra/polymarket-mcp-server
335TypeScript

polybot

Reverse-engineer strategies, find arbitrage on Polymarket.

git clone https://github.com/polybot-nexus/polybot
200Python

PredictOS

All-in-one multi-agent framework for prediction markets.

git clone https://github.com/PredictOS/predictos
200Python

dr-manhattan

CCXT for prediction markets with market-making support.

pip install dr-manhattan
181Python

polyterm

Polymarket in your terminal. Browse and trade from CLI.

cargo install polyterm
162Rust

CloddsBot

Autonomous AI agent trading across 1000+ markets.

git clone https://github.com/CloddsBot/cloddsbot
158Python

polymarket-pipeline

Event-driven AI news-to-trade pipeline for Polymarket.

git clone https://github.com/polymarket-pipeline/pipeline
154Python

gnosis/prediction-market-agent

AI agent that bets on prediction markets to benchmark reasoning.

git clone https://github.com/gnosis/prediction-market-agent
150Python

kalshi-trading-bot-cli

AI-native Kalshi CLI with Kelly sizing and risk management.

git clone https://github.com/kalshi-trading/bot-cli
149Python

kalshi-python

Official Kalshi Python SDK for market data and trading.

pip install kalshi-python
100Python

gnosis/pm-agent-tooling

Benchmark and deploy prediction market agents.

pip install prediction-market-agent-tooling
100Python

crm-cli

CLI customer relationship management. Contacts, deals, notes.

pip install crm-cli
72Python

AgentMail CLI

Email inboxes and transactional email for AI agents.

pip install agentmail
Python
Know one we're missing? Submit it →
Blog · 6

Latest from the blog

Insights on AI agents, prediction markets, and developer tools.

View all→

Automated Prediction Market Trading: CLI Agents on Kalshi

A practical guide for developers and traders on using CLI-based agents to automate order placement on Kalshi prediction markets. Covers thesis-driven trading logic, real tickers, and the agentic runtime behind production-grade automation.

Read→

Prediction Market Terminal Dashboard: Bloomberg-Style Monitoring for Kalshi Traders

A practical guide to building a professional-grade terminal dashboard for monitoring Kalshi prediction markets in real time. Covers CLI tooling, agentic scanning, position tracking, and thesis-driven trade execution.

Read→

Prediction Market Edge Detection: How to Find Mispriced Contracts on Kalshi

A systematic approach to finding mispriced prediction market contracts using causal models, orderbook analysis, and executable edge calculations.

Read→

Thesis-Driven Prediction Market Trading: Why Causal Models Beat Signal Chasing

Signal-based bots react to noise. Thesis-driven agents understand why prices should move. Here's how causal models change prediction market trading.

Read→

AI Agents for Prediction Markets: How SimpleFunctions Connects Claude to Kalshi

How to connect your AI agent to prediction market data using SimpleFunctions MCP server — get context, inject signals, and trade on Kalshi.

Read→

How to Build a Prediction Market Trading Bot with SimpleFunctions CLI

Build a prediction market bot that scans for edges, monitors thesis confidence, and executes trades on Kalshi — all from the terminal.

Read→