Everything we publish, free

Every Friday after the close: what moved the week, why it mattered, and what it did to the tape. Levels and scenarios, never calls.

• October 9, 2026 — a record S&P, and small caps lose the leadFor most of 2026 the small caps were the quiet winner. Last Friday IWM was still up 14.36% for the year against SPY's 12.86%, even after a September in which…10 min
• October 2, 2026 — 29,000 jobs, and the hike odds fellThe week turned on one number. The September jobs report showed payrolls up just 29,000, against a Dow Jones consensus of 84,000, and the unemployment rate…9 min
• September 25, 2026 — the 10-year at 5.2%, and chips ran anywayThe week's biggest move was not in stocks. The 10-year Treasury yield jumped more than 10 basis points on Thursday to as high as 5.223%, per…10 min
• September 18, 2026 — oil above $100, and the first hike since 2023The FOMC raised its target range by a quarter point on Wednesday, to 3.75%-4.00%. CNBC reported the vote at 12-0. Benzinga's coverage put it as the first…12 min
• September 11, 2026 — core CPI, and a $100 barrelFour sessions lower, one back, and an index that finished within a percent of where it started. SPY closed at 764.29, off 0.77% on the week. The number is…9 min
• September 4, 2026 — a payroll blowout, and the chips went upThe August employment report landed Friday morning at 162,000 jobs, against a Bloomberg-compiled consensus near 55,000. The unemployment rate held at 4.1%…10 min
• August 28, 2026 — Jackson Hole, and Nvidia's round tripThe week split in two. Wednesday night the largest company in the index reported a quarter that doubled its earnings year over year, and Thursday was a…11 min
• August 21, 2026 — the bond market wrote the scriptThe bond market wrote this week's script, and equities spent five sessions reacting to it. Yields on 30-year Treasuries pushed to their highest since 2007…13 min
• August 14, 2026 — cooler inflation, and the AI names sold anywayTwo inflation prints came in cooler than expected, the S&P 500 closed at what the coverage called a record on Thursday, and the two companies that sell the…11 min
• August 7, 2026 — a jobs miss, and the best week in monthsThe economy lost jobs in July, and stocks had their best week in months. Those two facts belong in the same sentence, because the second one happened because…11 min
• July 31, 2026 — a flat month and a correction, in the same four weeksJuly ended with the index you happened to own deciding your entire month. SPY +0.03%. QQQ -6.57%. IWM -3.08%. A flat month and a correction, describing the…10 min
• July 24, 2026 — two red weeks, and the 8-week line gave waySPY closed the week at 738.93, down 0.59%, its second red week in a row. Here's what the tape looks like going into an FOMC week.5 min

Our weekly macro and cross-asset outlook, out Sunday: the house view across four horizons, where it sits against consensus, a probability-weighted scenario set, and the one dated test that decides whether we were right. Every issue is graded in public the following week.

• October 11, 2026 — Record highs, a ten-year near 5.25%Tue 10/13: JPMorgan, Goldman Sachs, Citigroup, Wells Fargo. Wed 10/14: September CPI (headline polled at 3.6% against 3.4%), Morgan Stanley, Bank of America…21 min
• October 4, 2026 — The hike came off; the long end stayedThe week ahead is light on data and heavy on interpretation. ISM services Mon 10/5 (consensus 55.7), Fed minutes Wed 10/7, PepsiCo Thu 10/8, Delta Fri 10/9…22 min
• September 27, 2026 — Growth hot enough to buy an October hikeThe week ahead has two prints that set how much of an October hike gets locked in. August PCE Wed 9/30, core consensus +0.3% on the month; Micron reports the…23 min
• September 20, 2026 — A hiking cycle meets a supply shockThe week ahead carries no major data and ten Fed appearances, so the dots get interpreted, not tested. Flash PMIs Wed 9/23, jobless claims Thu, durable goods…25 min
• September 13, 2026 — The hike is priced; the 2027 dots are notWed Sept 16: FOMC decision 2:00pm ET with a new dot plot, presser 2:30. Futures put the quarter point near 85-90% (CME FedWatch). Lennar reports that…20 min
• September 6, 2026 — An oil shock the meeting cannot seeProducer prices Thursday, Oracle and Adobe after Thursday's close, August CPI Friday 9/11 with core consensus at +0.3% m/m and 3.1% y/y. All of it lands four…24 min
• August 30, 2026 — A stretched tape into a payrolls weekFour macro prints and one earnings report land before Friday: ISM manufacturing Tuesday, ADP and JOLTS Wednesday, ISM services Thursday, then August payrolls…23 min

How the instruments and the machinery actually behave — including the tests of our own ideas that failed.

• Agentic Trading Discipline: The Rules an AI Agent Trades ByAgentic trading discipline: rules enforced outside the model that decide when an AI agent may trade. What they must settle, and the rulebook Coil serves.10 min
• AI Trading Discipline: The System That Grades ItselfAI trading discipline is not willpower — it is externalized rules that survive a losing streak. The six mechanisms, each with a live working example.10 min
• Best AI Trading Bots 2026 — an Honest ReviewThe best AI trading bots of 2026, reviewed by a competitor with zero affiliate links: Trade Ideas, TrendSpider, Composer, QuantConnect, Coil, and more.15 min
• Best AI Trading Strategies, Measured — Not ListedThe best AI trading strategies, measured — not listed: we tested the five every list recommends and published the results, including the failures.12 min
• Can Claude trade stocks? Yes, wired to Robinhood — 5 rulesClaude cannot trade alone. Wired to Robinhood's MCP it places real orders. Which Claude client, what to permit, and the 5 rules to set before funding it.10 min
• Correction vs Bear Market: Why the Thresholds Are ArbitraryPullback, correction, bear market: what the thresholds are, who defines them, and why 19.8% and 20.2% are the same tape. The label is not a decision boundary.11 min
• Crypto trading signals for AI agents: BTC & ETH, benchmarkedCoil's crypto board scores BTC and ETH with long-only trend signals, backtest published beside buy-and-hold. It does not beat holding; it cuts the crashes.8 min
• Deterministic Trading Strategy for AI Agents: How to Run OneA deterministic trading strategy gives an AI agent fixed, tested rules: same inputs, same decisions. What it requires, and the long-only strategy Coil runs.9 min
• We tested betting on the decline. It did not work.We pre-registered a grid of inverse-ETF hedges and tested every cell against cash and a defensive macro sleeve. All 36 failed. The full result, published.11 min
• Does Chasing a Stock Near Its High Lose Money? A BacktestA backtest of buying a stock already up on the day and near its high: the question, the bar written first, the verdict and its limits. Coil Edge Research.2 min
• An MCP Server for Stock Scores: What Coil Gives an AI AgentCoil's MCP server gives AI agents scored stock data: regime, ranked buy list, per-stock reads, sector maps. Free one market day delayed; live by subscription.8 min
• Buying Pullbacks vs Chasing Breakouts: The Entry MathWhy traders buy leaders on pullbacks to support instead of chasing breakouts: the entry math, when breakouts do work, and how a rules engine encodes it.5 min
• Robinhood MCP Server Options: Official vs Trayd vs DIYThree ways to connect an AI agent to Robinhood: the official MCP connector, the open-source Trayd bridge, or your own, compared on auth, tools and orders.5 min
• Robinhood MCP Tools: The Complete Reference (All 53)Every tool on the Robinhood MCP surface, categorized: account, market data, orders, options, watchlists and scans — with the gotchas we measured live.7 min
• Self-tuning vs static trading botsWhat self-improving should and shouldn't mean in a trading bot: static params decay, unconstrained tuning overfits. The disciplined middle, with guardrails.9 min
• Single-Stock Leveraged ETF List: All 68 on Robinhood (2026)68 single-stock leveraged and inverse ETFs, every flag probed live against Robinhood's own endpoint: 19 block fractional buys, 17 have no 24/5 session.14 min
• SOXL Decay Calculator: 3x Daily Reset Math for Your HoldSOXL rebalances once a day, at the close — not continuously. Run your own holding period through the 3x decay calculator and see what an 80% fall costs you.16 min
• SOXS Is the Inverse of SOXL — Both Can Lose in Choppy TapeSOXS is the 3x inverse of SOXL on the same chip index: why both reset daily, why a flat choppy week can leave both lower, and why the inverse gaps overnight.12 min
• Glossary: automated & agentic trading termsPlain-English definitions of automated and agentic trading terms — AI agent, MCP, leveraged ETF, profit factor, drawdown, backtest, look-ahead bias, and more.12 min
• Why the Sentiment Gauges DisagreeAAII, Fear & Greed, the VIX and put/call ratios contradict each other because they measure different things on different clocks. What each one actually reads.11 min
• Why Your x402 Endpoint Is Unpayable by Half the Clientsx402 v1 clients parse the 402 body and want friendly network names; v2 reads the PAYMENT-REQUIRED header in CAIP-2. Serve one dialect and half can't pay you.7 min

Method, and where agentic trading is going.

• What Makes Software Agent-NativeAgent-native software ships machine-readable docs, manifests, deterministic CLIs, and verifiable claims. Here is what that looks like in practice.7 min
• Can AI Agents Buy Software Yet?How AI agents find and buy software in 2026: llms.txt, machine-readable manifests, and the checkout step that still needs a human.7 min
• AI Job Displacement and Market LeadershipAutomation waves reshape market leadership. Where earnings power migrates when AI changes work, and how rotation frameworks track it without forecasts.7 min
• AI Trading Scams: 9 Red Flags to Check FirstNine red flags to check before paying for any AI trading product: guaranteed returns, income promises, hidden risk, unverifiable track records and more.12 min
• Will Automation Give Better Returns Than Trading Manually?Automation does not add returns; it subtracts mistakes. What retail traders actually lose to, what automation fixes and what it can't, and how to judge one.8 min
• Buy Pullbacks, Not BreakoutsWhy disciplined systems buy leaders on pullbacks to real support instead of chasing breakouts. Entry price defines risk and structure gives a close stop.7 min
• Can an AI Agent Manage My Portfolio?Should an AI agent manage your portfolio? Why improvised trading fails, why rule execution is different, and the questions to ask any tool.7 min
• Can an AI Agent Trade Crypto in 2026? Broker by BrokerYes at Coinbase, announced at Robinhood. The broker-by-broker state of AI agents trading crypto in July 2026 — and where a benchmarked signal read fits.7 min
• Cash Is a PositionCash is an allocation, not the absence of an opinion. Why systems that must always be invested degrade, and why long-only ones need permission to sit out.6 min
• Crypto Trend Following vs Buy-and-Hold: BTC/ETH, HonestlyTen years of BTC/ETH through a 50/200-day trend rule: 59.6% vs 74.9% CAGR. It does not beat holding; it trades return for smaller crashes. Full table inside.7 min
• How to Read a Backtest Before You Trust ItA vetting guide for backtests: survivorship bias, fill assumptions, cost modeling, inflated t-stats, and the exact questions to ask any vendor.8 min
• Leadership Rotation, ExplainedMarket leadership rotates. Why last cycle's winners go stale, how relative strength ranks the new leaders, and why cash is a valid answer.6 min
• Leverage as an Accelerant, Not a DefaultLeverage works as a deliberate accelerant on high-conviction setups at reduced size, not a standing default. How Coil treats leveraged ETFs.7 min
• Leveraged ETF Gap Risk: When Stops Don't FireOvernight gaps skip stop losses. A 3x fund turns an index gap into roughly triple the damage at the open. How to score gap risk before entry, not after.6 min
• Market Weakness Is a Risk Input, Not a VetoWe built the rule that stops buying when the market is down, measured it over 2,649 S&P sessions, and threw it out. Weakness predicted drawdown, not returns.9 min
• MCP Explained for TradersWhat the Model Context Protocol is, what a broker MCP server exposes, why it beats screen-scraping and pasted API keys, and what agents still cannot do.9 min
• NVDL vs NVDA: What 2x Daily Really MeansNVDL targets 2x NVDA's daily return, not 2x the stock. How daily reset and compounding drift change multi-day holds, and when a system uses 2x.7 min
• Regime Concentration: Where Edge Actually LivesMost strategies earn their edge in specific regimes and roughly match markets elsewhere. Why vendors hide this and how to ask when a system fails.7 min
• Relative Strength vs Momentum: The Real DifferenceMomentum measures a stock's own trend. Relative strength ranks it against every peer. Why the difference matters and how a scanner uses both.7 min
• Robinhood's 70,000 Agentic Accounts, in ContextRobinhood says roughly 70,000 accounts have enabled agentic trading, well under 1 percent of its funded base. What that tells us about the era ahead.7 min
• Robinhood Agentic Crypto Trading: What's KnownRobinhood's Q2 report says customers can now trade crypto through AI agents. What shipped since July 10, what our agentic account can reach, what's unknown.11 min
• Robinhood Agentic Limited Margin: Trading Unsettled FundsRobinhood agentic accounts can switch to limited margin: trade with unsettled funds, no margin borrowing. What it fixes, and what to check before you flip.11 min
• Scheduled AI Agents: Software on a CadenceAgent scheduling lets trading software run on a cadence inside your AI agent instead of a server you maintain. What it enables, and what to demand.7 min
• Sector Rotation: A Practical GuideHow money rotates between sectors, how it shows up in relative strength before headlines, and why rotation informs entries but never dictates them.7 min
• Self-Custody Trading Software: Keep Your KeysWhat self-custody means for trading tools: why broker API keys should stay on your machine, the threat model of hosted platforms, and what to check first.7 min
• How Long to Hold a Single-Stock Leveraged ETF: 4 Red FlagsHow long you can hold one: days, not quarters. What the daily reset costs a multi-day hold in a single name, the gap no stop can catch, and four red flags.7 min
• Structural Stops vs ATR Stops, ExplainedWhy stops placed at the price where the thesis fails beat volatility-multiple stops, how wide stops interact with sizing, and why no stop is a guarantee.8 min
• Survivorship Bias in Backtests, ExplainedTesting today's index members backward quietly deletes every stock that died. How survivorship bias inflates backtests and how to fix it.7 min
• The Agent Economy: When Software Buys SoftwareAI agents are starting to discover, compare, and buy on our behalf. What that changes for builders and buyers, and where a $19/mo board fits.7 min
• TQQQ vs QQQ: The Long-Term MathTQQQ promises 3x the Nasdaq-100 per day, not per year. Plain-words math on daily resets, volatility decay in chop, gap risk, and holding-period honesty.7 min
• The Trading Bot Passive Income MythWhy passive income from a trading bot is a myth: adversarial markets, regime-dependent edges, overnight drawdowns, taxes, and what a bot honestly is.7 min
• Should You Turn Your Trading System Off During a Crash?A losing system and a broken system feel identical from inside a drawdown. The only way to tell them apart is with criteria you wrote down before it started.11 min
• UBI and AI in 2026: Where the Debate StandsThe UBI debate as AI displaces work: arguments for and against, what pilots suggest, and why people reach for ownership while policy stalls.7 min
• What Is Agentic Trading? A Plain ExplainerWhat agentic trading is, how it differs from robo-advisors and classic bots, and what a safe setup looks like. A plain explainer for 2026.7 min
• Why Long-Only Is a Feature, Not a LimitationShorting doubles the ways to be wrong and inverse ETFs decay on multi-day holds. Why Coil trades long-only, and what that choice honestly costs.7 min

Step by step, against a real brokerage account.

• How to automate SOXL/SOXS trading with an AI agentA practical, honest how-to: the components, the rules, and the validation it takes to run a hands-off SOXL/SOXS strategy through your own AI agent and broker.9 min
• Claude Code /loop for Robinhood: 3 ways to schedule itOne /loop 15m line, the standing instructions behind it, the Desktop-task and cron versions with the flags a headless run needs, and the hard end.13 min
• claude mcp add robinhood-trading: the flag it fails withoutclaude mcp add robinhood-trading --transport http https://agent.robinhood.com/mcp/trading — the line Robinhood publishes, and why the bare form fails silently.4 min
• Robinhood Agentic Trading with Claude: prompt, /loop, recordA copy-paste prompt for Claude, the /loop that keeps it re-checking through the US market day, and the engine's live record next to SPY, win or lose.23 min
• Robinhood agentic trading setup: why Claude won't tradeThe five-step setup, and the five reasons a connected agent still won't trade: the agentic account, the MCP line for Claude Code, the guardrails, the strategy.22 min
• Claude can't place trades on agentic Robinhood: 6 fixesConnected to agentic Robinhood but Claude won't place the order? Six symptoms in searchers' own words, each traced to its cause: account, confirm step, prompt.10 min

Coil against the alternatives, including where it loses.

• Coil vs ChartingLensCoil vs ChartingLens: an AI charting platform with a win-rate screener vs. a scanner that scores every S&P 500 and Nasdaq-100 name — and trades its board.12 min
• Coil vs Composer.tradeCoil vs Composer.trade: Composer is a hosted no-code strategy builder on subscription. Coil is a finished, long-only scored board, read in your browser or served to your own AI agent.9 min
• Coil vs Robinhood Agentic TradingRobinhood Agentic Trading is the free connection that lets an AI agent place orders. Coil is the long-only strategy that rides on it.8 min
• Coil vs StockHeroCoil vs StockHero: StockHero runs no-code trading bots on its cloud, subscription-priced. Coil is a long-only scored board that never holds your broker keys.11 min
• Coil vs Tickeron (AI trading)Coil vs Tickeron: Tickeron is a hosted AI trading subscription; Coil is a long-only scored board on subscription. Keys, transparency, breadth, cost.9 min
• Coil vs Trade IdeasCoil vs Trade Ideas: Trade Ideas leads AI day-trading signals on subscription. Coil is a long-only scored board, read in your browser or served to your own AI agent.11 min
• Coil vs Trading Bots & SignalsCoil vs trading bots and signal services, compared on keys, transparency, cost and who owns the risk. Coil is long-only, agent-native, and never holds your keys.9 min
• Coil vs Trayd MCPCoil vs Trayd (trayd-mcp): Trayd, a free open-source Robinhood-via-Claude bridge, needs your rules. Coil is a validated long-only scored board, rules built in.8 min
• Coil vs TrendSpiderCoil vs TrendSpider: best-in-class chart analysis on subscription, versus a long-only scored board your own agent can act on by rule, with the method written down and the record public.10 min

Also free

Not articles, but open to everyone and no account required.

• Trading glossaryPlain definitions for every term the board and the letter use.
• Every stock we scoreOne-day-delayed per-ticker reads across the whole scored universe.
• The live demoThe real dashboard on frozen data. No signup, no card.

About what we publish

What does Coil publish for free?

The Friday Wrap weekly market letter in full, every explainer and essay listed on this page, the trading glossary, one-day-delayed scores for every stock we track, and a live demo of the real dashboard on frozen data. None of it requires an account.

Do these pages give trading advice?

No. They explain how instruments and systems behave, and they report what our own tests found. Nothing here is investment advice, a recommendation, or tailored to anyone's circumstances, and the weekly letter states no forecasts, targets or trades.

Do you publish results that went against you?

Yes, and they are among the most-read pages here. The inverse-ETF hedge grid failed in all 36 cells and is published in full. The market-weakness gate we built turned out to be anti-predictive for returns and we threw it out in public. Where an outside source disagrees with our own data, we print both numbers.

How often does this change?

The Friday Wrap adds an issue every Friday. Explainers and essays are added as the work produces them, and this hub is generated from the published pages themselves, so it is never out of date with the site.