# AI Coders & Development Workflows

Working with Claude Code, Cursor, Codex, Gemini CLI. Tips, workflows, planning approaches for AI-assisted development.

## Blog Posts (70)

- [Jev Rank on TREMOR News Scores Every Headline Against What You Want to Read](https://www.tigzig.com/post/tremor-news-jev-rank-oct2026) - Tags: portfolio-quants, ai-coders
  TREMOR's news section now has Jev Rank, which scores about 800 headlines from 64 feeds from 0 to 100 against a reader's interests and sorts the page, in under 2 seconds. Three ways to use it: Run Jev Rank on default or saved interests, Edit interests with your own topics and example headlines, or a Focus box for a quick topic. Jev is TypeSafe's System One model; each headline is a yes or no question and the probability of yes is the score. Free, no signup.
  AI-readable: https://www.tigzig.com/ai/posts/tremor-news-jev-rank-oct2026.md
- [GLM-5.3 From Z.ai, Like Claude Mythos Preview, Can Build Cyber Exploits on Its Own, Anthropic Finds](https://www.tigzig.com/post/anthropic-glm53-open-weight-cyber-sep2026) - Tags: ai-coders, portfolio-quants
  Anthropic's Frontier Red Team tested GLM-5.3, an open-weight model from Zhipu AI, known outside China as Z.ai, and found it built a working Chrome exploit in 50 of 410 attempts against 56 for Claude Mythos Preview. NIST calls it the most cyber-capable open-weight model to date, about four months behind the best US models. Anthropic's tests show its refusals give way 64% of the time to a cover story and 100% once the refusals are edited out, which cost about 4,400 dollars of computing. Ten-page deck with the key findings and a link to the full report.
  AI-readable: https://www.tigzig.com/ai/posts/anthropic-glm53-open-weight-cyber-sep2026.md
- [Sophisticated Attacks No Longer Require Sophisticated Attackers. Anthropic's September Threat Intelligence Report, and What I See in My Own Logs.](https://www.tigzig.com/post/anthropic-threat-intelligence-cyber-sep2026) - Tags: security, ai-coders
  A 13-page deck on the cyber operations section of Anthropic's September 2026 threat intelligence report. A Russian espionage group left its AI rewriting malware until no security product could detect it, one operator downloaded 1.8 million phone apps to read the credentials developers had left inside them, and a third built a loop that reads a security product's code and keeps writing attacks until one works. The ways in are the familiar ones, stolen credentials, unpatched gateways, exposed services, SQL injection and phishing, and the difference is speed and scale: one break-in ran from a stolen developer login to full administrative control of a cloud in about three hours. Plus hotel guest wifi turned into a malware delivery system, and what has changed in my own edge logs since January.
  AI-readable: https://www.tigzig.com/ai/posts/anthropic-threat-intelligence-cyber-sep2026.md
- [Four Warnings About AI and Cyber Risk Arrived Inside Three Weeks. I Am Now Spending More Time Securing Than Building.](https://www.tigzig.com/post/four-warnings-ai-cyber-risk-sep2026) - Tags: security, ai-coders
  Four warnings on AI and cyber risk landed within three weeks: Google Threat Intelligence Group, a joint advisory from the NSA, CISA, FBI, Department of Energy and EPA, the Financial Stability Board chair's letter to the G20, and a departing OpenAI and Anthropic researcher. Set against nine months of running 40+ no-auth public tools behind Cloudflare, where volume and sophistication both rose. Nine-slide deck, each warning quoted from the document itself.
  AI-readable: https://www.tigzig.com/ai/posts/four-warnings-ai-cyber-risk-sep2026.md
- [Tigzig APIs Tell You What Went Wrong. If Your Code Reads Only the Status Code, You Never See It.](https://www.tigzig.com/post/tigzig-api-error-hints-read-the-body-sep2026) - Tags: database-ai, ai-coders
  Tigzig's free APIs return a tailored hint in the error body explaining what went wrong and how to fix it, but a client that reads only the status code never sees it. Shows how to pull the error block in Python, JavaScript and curl, why requests, httpx, aiohttp, urllib and raise_for_status all hide the body by default, and links the five API debugging guides you can hand to an AI agent.
  AI-readable: https://www.tigzig.com/ai/posts/tigzig-api-error-hints-read-the-body-sep2026.md
- [In Analytics the SQL Was Always the Easy Part. Anthropic's Data Team Put a Figure on It.](https://www.tigzig.com/post/anthropic-data-analytics-sep2026) - Tags: ai-coders, database-ai
  A read of two Anthropic posts on running self-service analytics with Claude, put into one deck with a practitioner's note at the end. Their headline numbers: without skills, accuracy on their own evals did not exceed 21 percent, and with skills it sits above 95 percent in aggregate on the same model and the same warehouse. Their offline accuracy then drifted from about 95 percent at launch to about 65 percent within a month, because the data model changed and the skill docs did not follow, which they fixed by colocating the markdown with the transformation models. The closing section covers what has worked in the author's own client work: a semantic layer holding the governed definitions, a separate skill per kind of work carrying its own judgment and tolerances, progressive disclosure so the agent does not read everything to answer one question, and continuous update as the thing doing most of the work.
  AI-readable: https://www.tigzig.com/ai/posts/anthropic-data-analytics-sep2026.md
- [AI Coders Are Not Just for Developers. An Anthropic Field Marketer Built His Own Weekly Sales Digest With Claude Code.](https://www.tigzig.com/post/anthropic-field-marketer-claude-code-aug2026) - Tags: ai-coders
  A read of Anthropic's post on Adam Ward, a field marketer with no engineering background who used Claude Code to build a Slack digest that reaches every sales rep on Monday morning, wired to BigQuery over MCP as the marketing team's source of truth. Covers the working habits that made it possible: briefing Claude like a new colleague, recording yourself thinking out loud and handing over the transcript, writing a fake version of the output first so there is a template to aim at, catching Claude's mistakes, starting small and piloting with a few users. Includes a ten-slide deck, plus pointers to Anthropic's marketing-ops and business-development write-ups, Boris Cherny on roles melting into five archetypes, and which AI coder you may already be paying for through a Google, ChatGPT or Claude subscription.
  AI-readable: https://www.tigzig.com/ai/posts/anthropic-field-marketer-claude-code-aug2026.md
- [A Free Markdown to PDF Converter Over API, With Typesetting Options. No Signup, No Key.](https://www.tigzig.com/post/markdown-to-pdf-api-typesetting-aug2026) - Tags: converters-tools, ai-coders
  A rebuilt markdown to PDF converter on Tigzig, free with no signup, no API key and no auth. Markdown can be pasted into the browser, POSTed to the API as JSON, or uploaded as a text or md file. Typesetting additions include explicit page breaks via a pagebreak directive, tables, nested lists, task lists and fenced code blocks, Greek letters, maths operators and arrows, currency signs from any script so a taka or rupee renders instead of an empty box, and emoji mapped to monochrome symbols with a small dot as the fallback. Also covers why an AI agent is usually better off rendering PDFs natively, and three tips for doing that at volume.
  AI-readable: https://www.tigzig.com/ai/posts/markdown-to-pdf-api-typesetting-aug2026.md
- [Reading Macro Charts on TREMOR? Send Them to the AI Agent Sitting There and Get a Read With the Latest News.](https://www.tigzig.com/post/tremor-ask-claude-agent-aug2026) - Tags: portfolio-quants, ai-coders
  An AI agent now sits inside the TREMOR macro dashboard and will read the charts in front of you, with three ways to ask. What changed this week sends every chart on the dashboard as one briefing, Analyze selected charts takes up to five and looks for trends, inflection points and correlations between them, and Deep research does the same while going out to newswires, ratings agencies and central bank releases and citing what it finds. The prompt is editable before it sends, output copies as markdown or prints to PDF, and past runs persist in browser storage. Each chart reaches the agent as its series, values, dates and source label, so it knows the provenance and how much history it has. Also the honest caveats: this is commentary rather than a research note, and the read is heavily shaped by the prompt, which is why the prompt is left customizable.
  AI-readable: https://www.tigzig.com/ai/posts/tremor-ask-claude-agent-aug2026.md
- [If a Tigzig API Keeps Returning the Same Error, Read the Response Body. There Is Almost Always a Hint in It.](https://www.tigzig.com/post/api-error-hints-aug2026) - Tags: database-ai, ai-coders
  Tigzig's APIs now try to work out what actually broke and put the answer in the error message, usually naming the exact character or the exact table, across the cricket database, the mutual fund and Yahoo Finance endpoints and others. The commonest failure on GET endpoints is not the SQL but the URL mangling it in transit, and the message will say so directly. The catch is that most tooling hides response bodies on an error: curl shows it unless -f is used, and exits 0 on a 400 so a shell chain carries on regardless; urllib raises an exception carrying nothing but the status; requests does not raise at all and leaves the hint sitting in r.text; and fetch never throws on a 400, so a try/catch around it catches nothing. Includes what to call in each case to actually see the hint.
  AI-readable: https://www.tigzig.com/ai/posts/api-error-hints-aug2026.md
- [Have Your AI Analyze the Cricket Data on Tigzig. 24+ Years of Ball by Ball, ODI and T20, Across 6000+ Matches.](https://www.tigzig.com/post/cricket-ask-your-ai-aug2026) - Tags: database-ai, ai-coders
  One pasted instruction is enough to put an AI agent onto 24 years of ball-by-ball cricket, ODI and T20 across more than six thousand matches, free and without signing up. Pointed at the machine-readable catalog at api.tigzig.com, the agent works out which service holds the cricket data, opens its OpenAPI spec, discovers the tables and columns, and writes the SQL itself. Three tables per engine joined on match_id, with match_info carrying venue, dates, teams, toss and result so a venue or toss cut needs no second source, and because the data is deliveries rather than pre-aggregated totals anything expressible in SQL is available. Works with Claude chat and Cowork, ChatGPT Work, and coding agents including Claude Code, Antigravity and Codex. Also the new cross-engine validation suite, which on its first run caught a single Test match present on one engine and missing from the other.
  AI-readable: https://www.tigzig.com/ai/posts/cricket-ask-your-ai-aug2026.md
- [Anthropic's Business Development Team Runs Inbound and Outbound With Claude Cowork. What the Post Actually Says.](https://www.tigzig.com/post/anthropic-bd-claude-cowork-aug2026) - Tags: ai-coders, chatgpt-integrations
  A summary deck and read of Anthropic's own post on how its business development team runs inbound and outbound with Claude Cowork, where roughly five hours a day of manual inbox and research work now goes into strategic work, and over a hundred accounts get researched overnight rather than one at a time. The four things the post calls out: build the knowledge base before the workflows, give Claude worked examples including each rep's own writing style, keep a person on every send, and write corrections back into the skill so a mistake is recorded once. Plus what actually comes with a Claude or ChatGPT subscription, when to use Claude in Excel over Cowork, and links to Anthropic's other practical guides including the argument for HTML over Markdown as an output format.
  AI-readable: https://www.tigzig.com/ai/posts/anthropic-bd-claude-cowork-aug2026.md
- [ChatGPT Work Comes With Your ChatGPT Go or Pro Subscription. Here Is What It Can Do.](https://www.tigzig.com/post/chatgpt-work-ai-worker-analytics-aug2026) - Tags: ai-coders, chatgpt-integrations
  ChatGPT Work is included with a ChatGPT Go or Pro subscription and plays the same role as Claude Cowork, reading CSVs and parquet files, running analysis, building models, tools, automations, reports and presentations, and calling external APIs like FRED or the free Tigzig ones to pull data. A bit clunky next to Cowork, but only weeks old. Three copy-and-run prompts to try it with: a FRED macro report rendered to portrait HTML pages and converted to PDF, an instruction to go find the QRep comparison API unaided and build an interactive HTML risk tool from 81 metrics across six large-cap names against SPY, and a 168 MB parquet of live India mutual fund NAV data with 37 million records to explore from scratch.
  AI-readable: https://www.tigzig.com/ai/posts/chatgpt-work-ai-worker-analytics-aug2026.md
- [If You Pay Rs 1,950 a Month to Google, You Already Have Antigravity, a Full AI Coding Agent.](https://www.tigzig.com/post/google-antigravity-ai-coding-agent-aug2026) - Tags: ai-coders
  Antigravity is Google's agentic coding tool, and anyone on the Rs 1,950 a month Google subscription already has access to it at raised limits, with a free tier available for trying it first. Because it is a full coding agent rather than a chat window, it manages local files, pulls data straight from feeds like Yahoo Finance or FRED, builds custom HTML dashboards, and helps with Excel automation through VBA, Power Query M code or Python in Excel via xlwings Lite. The developer-style interface can be softened into something closer to a chat app by hiding the file panels and switching to a light theme. It runs on your own machine with terminal access, so the per-command approval prompt should stay on. A note on where it sits against Claude Code, which remains the primary tool here on quality.
  AI-readable: https://www.tigzig.com/ai/posts/google-antigravity-ai-coding-agent-aug2026.md
- [Technical Analysis on Any Ticker Without Visiting My Site. Just Ask Your AI Agent.](https://www.tigzig.com/post/tigzig-technical-analysis-ask-your-ai-aug2026) - Tags: technical-analysis, mcp, ai-coders
  The TigZig technical analysis API produces a formatted PDF and web report with charts, indicator readings and an outlook for any Yahoo Finance ticker, covering stocks, indices, FX, crypto and commodities, and it can be reached entirely through an AI agent without opening the site: paste a plain-English request naming the ticker and the tigzig API, and the agent discovers the API catalog, picks the right endpoint and reads the rate limits itself. Model choice is available on request. How it is built: prices from Yahoo Finance on daily and weekly timeframes, indicators computed with finta (EMAs, MACD, RSI, Bollinger Bands), two charts drawn with matplotlib, then both the numbers and the chart images passed to a vision model that reads the chart and writes the interpretation, with the returned markdown converted to PDF and HTML by a second ReportLab-based service. One FastAPI app serves the web front end, the REST endpoint and the MCP endpoint, so a browser and an agent execute the same code, and the whole thing is open source for anyone wanting to self-host or customise it. Closes with an explicit scope statement: this is a reporting tool rather than investment advice or research, models can misread a chart confidently, and the author's own practice is to use AI for the heavy lift and alternative angles while validating every deterministic output, the numbers and formulas, against more than one source.
  AI-readable: https://www.tigzig.com/ai/posts/tigzig-technical-analysis-ask-your-ai-aug2026.md
- [Comparing Securities on Returns, Risk and Drawdowns. Four Routes into QuantStats, Depending on What Your AI Can Do.](https://www.tigzig.com/post/quantstats-four-routes-qrep-agent-aug2026) - Tags: portfolio-analytics, mcp, ai-coders
  Four ways to get portfolio and security analytics out of QuantStats, the open-source library by Ran Aroussi who also wrote yfinance, with the right route depending on what your AI can actually do. An AI chat such as ChatGPT cannot install Python, so it connects through an MCP server, and the open-source QREP MCP server is one option. An AI agent such as Claude Cowork, Claude in Excel or PowerPoint, or ChatGPT Work, installs the package itself, runs the numbers and formats the output or builds a dashboard, and can equally call the QREP API directly for a fixed set of 81 metrics with no setup, or connect over MCP. For interactive work on screen, the QREP browser tool runs the cuts and generates a QuantStats-format tear sheet. Tool and API are both QuantStats-powered and both open source, with a few QREP implementation choices that shift a handful of numbers slightly, documented in the methodology page. Also notes that an agent with Python can be asked to write the metric functions from scratch, which is useful for validation and for metrics such as rolling returns. Adds a clarification on what a chat app can actually run in its Python sandbox, where ChatGPT is more restricted than Claude.ai and the desktop cowork apps are more capable again, with the practical advice being to ask the assistant what it can do and test it. Includes copy-paste agent prompts for both the run-it-yourself and the API route, using the Mag 7 against the S&P 500 over five years, plus practical notes for the browser tool: ticker search rather than typing, CSV export of the metrics table and raw prices, adjustable omega threshold and var/cvar confidence and tail cutoff, and browser-computed EMA, RSI and MACD technicals with the formulas in the docs.
  AI-readable: https://www.tigzig.com/ai/posts/quantstats-four-routes-qrep-agent-aug2026.md
- [Claude Cowork Runs a Real Python Sandbox. I Put 37 Million Rows of Mutual Fund NAV Data Through It.](https://www.tigzig.com/post/claude-cowork-python-sandbox-mf-data-aug2026) - Tags: ai-coders, mutual-funds, duckdb
  What Claude Cowork's Python sandbox can actually do for analytics work, tested on a real dataset. It ships with pandas, numpy, scipy, duckdb, matplotlib and yfinance, installs scikit-learn and statsmodels in under a minute, and takes polars and pyarrow too, though packages needing compilation can time out on install (xgboost did). An approximate benchmark run inside the sandbox reports two cores of an AMD Ryzen 7 7730U, about 4 GB of memory and 10 GB of SSD-class disk. The worked case: downloading the free India mutual fund NAV dataset published on tigzig.com, a 175 MB parquet of 37 million-plus rows going back to April 2006 plus the scheme master, and running category-level risk and return analytics with results back in under a minute per iteration. The same Python sandbox is available in Claude for Excel and PowerPoint. Includes a copy-paste prompt for pointing any agent at the same live datasets, with the four things that trip people up on this data: matured and merged schemes are retained so it is survivorship-bias free, a row is a plan-and-option variant rather than a fund (about 8,500 codes publishing daily, roughly 1,950 funds as AMFI counts them), and scheme names and categories shift under the SEBI recategorisation so joins should key on scheme code or ISIN. Closes with why DuckDB and Parquet are the working default: the same NAV data is 3.8 GB as plain CSV against about 175 MB as parquet, columnar storage means a query reads only the columns it touches, and DuckDB's statistical functions run in SQL without dropping into pandas.
  AI-readable: https://www.tigzig.com/ai/posts/claude-cowork-python-sandbox-mf-data-aug2026.md
- [Tigzig is Built AI-Agent First. 300+ Assets, and the Fastest Way In Is to Ask Your AI.](https://www.tigzig.com/post/tigzig-agent-first-ask-your-ai-aug2026) - Tags: ai-coders, infrastructure
  Tigzig is built AI-agent first, so the quickest route into the site is to ask an assistant rather than browse it. All 300+ assets, guides, live apps, repos, open APIs and MCP servers, are organised to be machine readable and agent accessible, with worked example prompts for ChatGPT, Claude, Gemini or Meta AI such as asking what the site has for building an Excel tool, whether it carries analysis on Donald Trump's trades, or to pull the author's posts on economy and markets. An AI coder (Claude Cowork, ChatGPT Work, Claude Code, Codex) can go further and fetch the code, call the APIs and deploy and run something for you. Also covers two changes in traffic: a fair portion of API calls now arrive from agents and automated tools, so some callers are meeting rate limits they would not have hit before, with limits under active adjustment and the limits themselves being documented across response headers, 429 bodies, the machine-readable API roots and the human docs, starting with the MFPRO API and extending to other high-usage APIs including the Yahoo Finance and FFN portfolio tools. Suggested workarounds are batching identifiers into a single call where an API supports multiple tickers or scheme codes, or taking the whole database as a file for TREMOR and the mutual fund API. Ends on where the traffic now comes from: ChatGPT is the number one referrer after Google, and API callers arriving as ChatGPT and Claude send standard headers and behave in a predictable, systematic way.
  AI-readable: https://www.tigzig.com/ai/posts/tigzig-agent-first-ask-your-ai-aug2026.md
- [Pulling Company Financials Straight From SEC Filings. Free, No Vendor, No API Key. Three Routes Tested Head to Head.](https://www.tigzig.com/post/sec-financials-three-ways-xbrl-edgartools-jul2026) - Tags: portfolio-quants, ai-coders
  Three routes to company financials from SEC filings, tested head to head in a Claude Cowork session on the same seven companies and six quarters: the SEC XBRL API directly, the edgartools Python library, and yfinance. All three agreed on the numbers; they differ on coverage and traceability. Verdict: edgartools first - it ships a statement API that pulls a full statement in one shot with indent level and parent, a customizable tag-standardization layer, de-cumulation, EDGAR full-text search, 13F institutional holdings, TTM calculations and segment facts, plus Claude skills and an MCP server. The raw SEC API stays for raw SEC API for cross-sectional screening of thousands of filers and as an independent second path - a check that earned its keep when the one disagreeing figure turned out to be a mishandled restatement. Links to a full 17-section write-up covering XBRL concepts/tags/facts, the four API routes as slices of a cube, frames coverage, taxonomies, saving filings, R-pages, standardisation, duration vs instant, the yfinance reconciliation, edgartools capabilities and its shipped Claude skills, and limits - with every Python script, JSON dataset and source filing published alongside so any number can be traced to the code that produced it. edgartools is by Dwight Gunning, MIT licensed, 2,000+ stars, ships Claude skills and an MCP server.
  AI-readable: https://www.tigzig.com/ai/posts/sec-financials-three-ways-xbrl-edgartools-jul2026.md
- [TigZig is Built Agent First. 300+ Assets on AI for Analytics. Just Ask Your ChatGPT, Gemini or Claude.](https://www.tigzig.com/post/tigzig-agent-first-ask-your-ai-jul2026) - Tags: ai-coders
  Release note explaining that tigzig.com is built agent-first: 300+ assets (210+ guides, 40+ live apps across 13 categories, open-source repos, a free open API and live MCP servers) on AI for analytics, and the recommended way in is to ask your own AI - ChatGPT, Gemini, Meta AI on WhatsApp, Claude, or an AI coder like Claude Code / Cursor / Codex - rather than browse. Includes example prompts (market valuations, database AI, macro data, app security, cloning and deploying the open-source repos), quick links to the main hubs (markets, claude-for-analytics, mutual-funds, mcp-landing, analysis), and pointers to the two explainers on chat apps vs cowork apps vs embedded apps vs AI coders. Single-image release post with the guides-and-tools one-pager PNG at the bottom.
  AI-readable: https://www.tigzig.com/ai/posts/tigzig-agent-first-ask-your-ai-jul2026.md
- [Jamie Dimon Uses Claude Code to Generate Dashboards. Are You? You Need an AI Worker or AI Coder, Not an AI Chatter.](https://www.tigzig.com/post/ai-worker-not-ai-chatter-jul2026) - Tags: ai-coders
  Practitioner commentary framed around Jamie Dimon (JPMorgan CEO) using Claude Code to generate dashboards - and the author's argument that for serious CXO or analyst work you need an AI worker or an AI coder, not an AI chatter. Three-tier taxonomy: (1) AI Chatter = ChatGPT / Claude / Gemini in browser - fine for questions, can generate code and do Python, but no longer enough on its own; (2) AI Coworker = Claude Cowork and recently launched ChatGPT Work - can connect to external systems, pull data, run analysis and models, build decks, automate - a full working assistant that can check the work your team sends over or build rapid prototypes instead of lengthy team calls; (3) AI Coder = Claude Code / Codex / Cursor - all of the above plus complex models, scrapes, full-stack apps from build to deployment, essentially a data-science + research + tech team combined. How-to-start section: install the Claude or ChatGPT desktop app on your current plan, try things out, share the author's Cowork PDF guide with the AI itself and have it tutor you with test cases (they are great tutors, they just need to be asked), and lean on the tutorials from Anthropic / OpenAI plus YouTube. Cost economics: both Claude and ChatGPT start at $20 (light usage only), serious work is $100, heavy work is $200; ChatGPT has a free tier plus an $8 plan; team plans up to 150 users on flat per-seat pricing with usage included and the premium seat covering Claude Code and Cowork. Author's own economics as a freelancer: Claude Code is the primary tool from ML models to full-stack apps, ROI is a no-brainer, and 2 business clients moved to Claude Code with 'game-changer' impact in their own words. Enterprise tier is a different beast - a seat fee plus usage billed at API rates (not the flat team pricing), and the same heavy usage that fits inside the subsidized $200 individual plan can easily run $8K to $15K per person per month at metered API rates if pushed hard. Closes with a set of questions the AI Worker hub answers (Claude in Excel + PowerPoint, what works and what doesn't, MCP Server explained, how to connect to external data) and a link to the hub. Single-image analytical post with a one-pager PNG at the bottom.
  AI-readable: https://www.tigzig.com/ai/posts/ai-worker-not-ai-chatter-jul2026.md
- [AI Is Not Just Chat. Cowork Apps Are Where the Action Happens.](https://www.tigzig.com/post/ai-is-not-just-chat-cowork-jul2026) - Tags: ai-coders
  Practitioner commentary on where the real AI work happens for analyst / builder workflows - the Cowork apps (Claude Cowork and now ChatGPT Work). These are the desktop apps that come as part of your Claude / ChatGPT subscription, and they do not just chat - they run code, pull data from external sources and CSVs, analyze it, and build reports and decks - basically do the work. Post plainly points out this is not one-golden-prompt magic - you still need to know what you are doing, provide context, guide, iterate, and most importantly validate results; the buck stops with you and the apps are just tools. Links two go-deeper resources: a 23-page guide on what Claude Cowork can do (author has just tried ChatGPT Work and confirms it can do the same things) and the Claude for Analytics hub. Quick-start tips section: install the desktop app and try things; if you want structured onboarding, share the guide PDF with the Cowork app and have it tutor you with test cases; the Anthropic docs at claude.com/resources/tutorials are simple language + good examples, and the author's favourites are curated in the Analytics hub. Second half - practitioner notes on when to use the other Claude flavors instead. Claude in Excel: stick to it for Excel work, far more efficient than doing Excel through Cowork. Claude in PowerPoint: efficient for PPT, but lately the author has stopped using PowerPoint - prepares decks in HTML from Cowork or Claude Code and converts to PDF, because HTML-to-PPT is wonky on font and placement and needs iterations. Claude Code: the author's bread and butter - the 40+ apps on tigzig.com have been built, maintained, and are defended by Claude Code (earlier Cursor); also solid for ML models with diagnostics reviewed as HTML reports, though the author tends to prefer CC over Cowork for that. Claude Design: rarely used - Claude Code or Cowork are faster for report and deck templates, Claude Design is overkill.
  AI-readable: https://www.tigzig.com/ai/posts/ai-is-not-just-chat-cowork-jul2026.md
- [Anthropic's Marketing Ops on Using Claude for Campaign Reporting - and How I Do the Same for Small Businesses. Notes on Skills.](https://www.tigzig.com/post/anthropic-marketing-ops-claude-skills-jul2026) - Tags: ai-coders
  Commentary on Anthropic's marketing-ops team post about how they run campaign reporting and builds using Claude Cowork - the practical details of moving from chat to actual coworker runs. Anthropic's setup, per their public post: weekly metrics reporting cut from days to hours using Sunday runs, and campaign ops triggered via Slack, routed to specialists, and verified by an audit agent - everything controlled via skills. Post carries an 8-page summary deck of the Anthropic setup, and links to Anthropic's own full post. Second half is the author's own practical experience running campaign ops for small businesses (no Slack on the client side - it is calls, emails and WhatsApp) - the core setup revolves around four skills (list creation, list validation, response modeling, campaign measurement), each carrying business-specific context and business rules very specific to the domain and client, plus pointers to baseline Python scripts that the agent modifies. Modeling / measurement skill produces the outputs the author wants (validations, decile tables, Gini, ROC charts) - nearly all outputs as HTML nowadays. Measurement skill also includes business rules like what counts as an attributable conversion. Important tips section - every recurring task = skill; skills carry business context and how-you-want-it-done, not code (agents don't need to learn Python; they need to learn how you want it done); create baseline scripts and point at them from the skill; update skills immediately after every task and session; and every few weeks run a skills reorg (author has a skill for that too). Worth-reading section links Anthropic's own Complete Guide to Building Skills for Claude PDF, their agents-and-tools docs, and the author's Claude Cowork guide + Claude for Analytics hub. Closes with a section of the author's own tips on skills: names carry more weight than you'd expect (renaming working-style.txt to amar-user-profile.txt killed a persistent ambiguity); skills vs CLAUDE.md - any 'when this comes up, do these 5 steps' workflow belongs in a skill, the skill description IS the pointer, not a CLAUDE.md doc line; live code pointers over prose - agents open the real script; description does the triggering, body does the doing (use YOUR actual phrasing in the description - author's own examples: 'do the cache warmer thing', 'bat it', 'break it', 'do the SEO thing').
  AI-readable: https://www.tigzig.com/ai/posts/anthropic-marketing-ops-claude-skills-jul2026.md
- [FRED, 845,000 US Economic Series - Your AI Can Pull Any of Them Directly. No MCP, No Plugin, No Install.](https://www.tigzig.com/post/fred-your-ai-pulls-directly-jul2026) - Tags: ai-coders, infrastructure
  Pitch + practical starter for pointing an AI at public data APIs directly, no plugin required. Headline: FRED, the St. Louis Fed's free database, has ~845,000 economic time series (inflation, jobs, GDP, rates, markets) and your AI (Claude Cowork, OpenAI Codex, Claude in Excel and PowerPoint) can just reach in and pull any of them - no MCP server, no plugin, no install, free. Mechanism: each of those AI tools has a Python sandbox inside, so calling an API is just Python for them, and an API reads like plain English ('get me US gasoline prices from FRED, chart it, tell me what changed' - it goes, connects, pulls, analyses, pushes into Excel or a slide or an HTML report). Three-step get-going: grab a free FRED API key at fredaccount.stlouisfed.org/apikeys (generous rate limits), first time enable outside access for the agent (one setting), then just ask. Post also links two go-deeper resources: the Claude Cowork guide (a deck on how the author runs the whole thing) and the Claude for Analytics hub. Second half lists a handful of free APIs the author keeps using: FRED (845k US series, key required), World Bank (api.worldbank.org, no key, global GDP + inflation + debt), US Treasury FiscalData (api.fiscaldata.treasury.gov, no key, federal debt + rates), ECB reference FX (api.frankfurter.dev, no key), Yahoo Finance (via yfinance in the sandbox). Plus author's own free tigzig.com endpoints (all on api.tigzig.com, no key): Tremor (curated US + global macro/credit/valuation indicators), Mutual Fund NAV India from AMFI (19.8M records, 17k schemes), Vigil (India corporate red flags), Portfolio Analysis, AI Technical Analysis, Securities Comparisons. For a human - tigzig.com API/MCP menu. For an agent - copy the page or point it at api.tigzig.com. Single-image release post.
  AI-readable: https://www.tigzig.com/ai/posts/fred-your-ai-pulls-directly-jul2026.md
- [Copy Page as Markdown - Every Page on Tigzig.com Now Hands Your AI Full Context in One Click.](https://www.tigzig.com/post/tigzig-copy-page-button-jul2026) - Tags: ai-coders, converters-tools
  Release note announcing the site-wide 'Copy page' button on tigzig.com. Every page (API doc, MCP guide, blog post, hub, landing page) now carries a Copy page split-button that hands the reader's AI (Claude Cowork / Codex / ChatGPT / Gemini) the entire page as clean markdown - full context in one paste, no partial copy paste and no screenshots. Workflow: read an API or Excel-MCP guide, click Copy page, paste into your AI, and it takes it from there - explains the endpoints, writes the code, deploys on the user's machine. Alternative flow for the browse-averse: skip the site entirely and just ask the AI directly ('go to tigzig.com and find what to do for my Excel to talk to my database.. and help me deploy it also'). Working example inside the post: pasted the MF NAV API page (tigzig.com/apis/mf-nav) into Claude Cowork with a 5-year SIP analysis prompt - it searched the scheme by name, pulled 2,464 daily NAVs, ran 120 monthly installments of Rs 10,000 on Parag Parikh Flexi Cap Direct, and returned Rs 12 lakh grown to Rs 30.5 lakh, XIRR 17.75% - one paste, no code from the user. Framing: the whole site is now AI-Agent-First, humans welcome too. Standalone one-pager PNG attached at the end.
  AI-readable: https://www.tigzig.com/ai/posts/tigzig-copy-page-button-jul2026.md
- [Boris Cherny: Engineering, Product, Design, Data Science Are Melting Into Archetypes. Seeing the Same.](https://www.tigzig.com/post/boris-cherny-archetypes-roles-melting-jun2026) - Tags: ai-coders
  Commentary on Boris Cherny's (Creator & Head of Claude Code at Anthropic) point that engineering, product, design and data science are melting into one kind of role and the work now maps to archetypes (prototyper, builder, sweeper, grower, maintainer) rather than to titles. The author sees the same in small businesses where he is bringing in Claude Code and Claude Cowork. Day-to-day pattern: marketing and finance folks building their own local automations and dashboards (often in VBA and Python); developers and builders turning those into production-grade tools, then maintaining them, working full-stack regardless of where they started out (frontend, backend, etc.); data scientists running more models, better validated, refreshed more often, and creating working prototypes of deployable tools and handing off to builders. Boris' team-mix view: pre-PMF products lean on archetypes 1+2+3, growing/found-PMF products on 2+3+4 plus some 5, strong-PMF products on 3+4+5 plus some 2. The 4-slide carousel pairs Boris' original framing with the author's own field observations. Work moves across the team; job titles matter less and less.
  AI-readable: https://www.tigzig.com/ai/posts/boris-cherny-archetypes-roles-melting-jun2026.md
- [Stop Browsing Tigzig.com. Ask ChatGPT First. (Or Whichever AI You Use.)](https://www.tigzig.com/post/tigzig-ask-chatgpt-stop-browsing-jun2026) - Tags: ai-coders
  Standalone one-pager release: tigzig.com now has 300+ assets (live apps, repos, build guides, data APIs, MCP servers, analysis) and is indexed across ChatGPT, Claude, Gemini and Meta AI on WhatsApp. Don't click through it the old way - just ask whichever AI you keep open, and it walks you through what is there. Sample agent prompts in the deck: 'tigzig.com - anything on Python in Excel?', 'tigzig.com - anything on Claude in Excel?', 'tigzig.com - any guides on security for AI tools?', 'tigzig.com - any portfolio analysis tools I can use?', 'tigzig.com - pull the Claude Cowork guide', 'tigzig.com - what is on mutual fund analytics?', 'tigzig.com - any infra guides for tool builders?'. Faster than clicking around, and a fair way to sanity-check the site too. All free, AI-agent-first.
  AI-readable: https://www.tigzig.com/ai/posts/tigzig-ask-chatgpt-stop-browsing-jun2026.md
- [Claude Cowork and Claude Code Have Quietly Become My Universal Front End. More Than Excel, Jupyter or Colab.](https://www.tigzig.com/post/working-with-claude-cowork-claude-code-jun2026) - Tags: ai-coders, mcp
  First-person guide on running almost the whole analysis-and-modeling pipeline from one place. For the author, Claude Cowork (and Claude Code) have quietly become the universal front end - past Excel, Jupyter and Colab - covering exploration, cleaning, ensembles, hyperparameter tuning, validation, and final presentation deck. Each iteration is shared as HTML reviewed in the browser, so a session ends with both a model AND full documentation written as work progressed. Same flow works for database-connected analysis, not just model builds. Excel work goes through Claude in Excel or Cowork; PowerPoint too. Full-stack builds stay on Claude Code. The difficult part was letting go: after years of notebooks and code editors, with the comfort of code you can touch and feel, stepping back to talk into a chat box and review the results took some getting used to. Once you let go, it is way faster and gives much better results, since the human focuses on validations, logic, business and domain pieces - the places where an AI agent still needs guidance. Includes a 23-slide deck that Cowork put together about itself: chat -> coworker shift, desktop + Code teammate, hitting APIs (FRED, Yahoo Finance), reaching backends via MCP, working files in place, cleaning messy spreadsheets, 11.85M-row / 1.6GB analyses in 13 seconds, EDA-as-browser-page iteration loop, gradient-boosting model builds with ROC / gains, A/B significance tests, packaging as HTML/PDF/Word/PPT/Excel, live dashboards that pull fresh data, real URL fetching (not snippets), scheduled briefings, browser/desktop control, cross-session memory, sandbox toolkit (4GB RAM / 10GB disk / pandas / polars / pyarrow / scikit-learn / matplotlib / OpenCV / Pillow), no-code-on-your-machine handoff to Claude Code, and Anthropic's own resources (Cowork product, best-practices guide, the HTML-beats-Markdown post, computer-use docs). Deck self-built in Cowork as HTML then exported to PDF.
  AI-readable: https://www.tigzig.com/ai/posts/working-with-claude-cowork-claude-code-jun2026.md
- [AI Is the Front Door Now. Don't Dig Through Tigzig.com - Ask Your AI About It.](https://www.tigzig.com/post/tigzig-ask-your-ai-front-door-jun2026) - Tags: ai-coders, mcp
  Standalone one-pager release: AI is the front door now. Just like you ask your AI about Cloudflare, Vercel or AWS instead of digging through the docs, tigzig.com (40+ live apps, repos, ~200 build guides, data APIs, MCP servers, 100s of assets) is indexed across the major AI assistants - ChatGPT, Gemini, Google AI, Claude, WhatsApp Meta AI - so you can ask whichever one you keep open what's there, what's not, and how to use it before you even visit. Sample questions: Is tigzig.com legit? What can it do with my database? Does it have a security checklist for AI tools? How do I pull a portfolio or technical-analysis report? What free data APIs does it have? Quicker than browsing, and a fair way to sanity-check it.
  AI-readable: https://www.tigzig.com/ai/posts/tigzig-ask-your-ai-front-door-jun2026.md
- [Agentic Coding and Persistent Returns to Expertise - Anthropic's 400k Claude Code Sessions, Summarized](https://www.tigzig.com/post/anthropic-claude-code-expertise-survey-jun2026) - Tags: portfolio-quants
  Anthropic just published 'Agentic coding and persistent returns to expertise' - a privacy-preserving analysis of ~400,000 interactive Claude Code sessions from ~235,000 people, October 2025 to April 2026. Three takeaways pulled into a 9-slide summary deck with every headline a direct quotation: (1) coding background is becoming less relevant - every major occupation succeeds at rates within a few points of software-related occupations (software engineers verified at 34%; all ten largest groups within seven points); (2) management occupations show the highest verified success, slightly above software engineering, which Anthropic reads as management skills transferring to directing an agent (with a measurement caveat); (3) the more domain expertise a person brings, the more work Claude does per instruction (more than twice the actions, five times the output from the same prompt) and the more often the session ends in success (verified success doubles 15% to 33%, novice-to-intermediate is where most of the jump happens, and only 5-7% of expert sessions end abandoned vs 19% for novices). Bonus: the share of sessions spent fixing broken code fell from 33% to 19% while average task value rose 27% vs a freelance-market benchmark. Anthropic's honest caveat: classifiers remain challenging to validate at scale; they cannot measure whether the code is actually used. Matches the author's 25 years as an analyst - on coding the agents win by multiples, on deep domain the human still wins.
  AI-readable: https://www.tigzig.com/ai/posts/anthropic-claude-code-expertise-survey-jun2026.md
- [Claude Code Was Getting Dumber. Stopped It from Auto-Spawning Sub-Agents. Performance Came Back Instantly.](https://www.tigzig.com/post/claude-code-sub-agents-performance-fix) - Tags: ai-coders
  Claude Code Opus 4.6 was auto-spawning sub-agents for every task causing rework loops, sloppy work, and lost context. Sub-agents reported Done without validation. Fixed with one CLAUDE.md global rule: no agents without express authorization. Performance returned instantly. Still uses sub-agents for bulk tasks but main agent sets up process first.
  AI-readable: https://www.tigzig.com/ai/posts/claude-code-sub-agents-performance-fix.md
- [Anthropic Released Claude Managed Agents Yesterday, Live on TREMOR Today](https://www.tigzig.com/post/tremor-claude-managed-agents-macro-analysis) - Tags: portfolio-quants, ai-coders
  Claude Managed Agents integrated into TREMOR macro dashboard. AI analysis on 70+ indicators across US and India. Three modes: full dashboard briefing, selected chart analysis, deep research with citations. Covers equities, bonds, gold, oil, forex, unemployment, delinquencies, valuations, inflation.
  AI-readable: https://www.tigzig.com/ai/posts/tremor-claude-managed-agents-macro-analysis.md
- [Claude - The Universal Agent](https://www.tigzig.com/post/claude-the-universal-agent) - Tags: ai-coders, claude-in-excel, mcp-servers
  Claude went from AI coder to universal agent in 6 months. Excel formulas, slide decks, PDF reports, MCP connectors, database queries, security scans, full-stack builds - one tool handles what used to need 5-6 separate tools. 10-slide overview covering the evolution and use cases across roles.
  AI-readable: https://www.tigzig.com/ai/posts/claude-the-universal-agent.md
- [Claude the Hunter-Killer - Have You Seen Your Nice Little Claude Run a Penetration Test on Your Apps?](https://www.tigzig.com/post/claude-the-hunter-killer-pen-test) - Tags: security, ai-coders, infrastructure
  Real-world penetration test using Claude Code against a hardened DuckDB dashboard app (230M rows, IMDB data). Despite API keys, Cloudflare edge rate limiting, JS challenge, SQL blocklist and backend rate limits, Claude found repeat() memory bombs that finish within timeout, metadata leaks, and missing conn.interrupt() leaving DuckDB crunching after timeout. Shows how Playwright bypasses JS challenge using real Chrome and fires attacks from same-origin context. Practical lesson: use separate Claude instances for coding and pen testing.
  AI-readable: https://www.tigzig.com/ai/posts/claude-the-hunter-killer-pen-test.md
- [Claude in Excel + MCP + xlwings Lite + Claude Code: Pull Data, Build a Model, Validate It.](https://www.tigzig.com/post/claude-in-excel-mcp-xlwings-lite-claude-code-combining-4-tools) - Tags: claude-in-excel, mcp, xlwings-lite, ai-coders, portfolio-analytics
  Live walkthrough of S&P 500 forward returns scenario model built with four tools. Claude in Excel pulls data via YFIN MCP server and builds MAP/LAMBDA formula model. Claude Code validates offline with independent Python recomputation - catches a formula error that in-Excel checks missed. xlwings Lite generates distribution charts (fan, ridgeline, raincloud) from Python into Excel. Covers when to use each tool and MCP architecture for client work.
  AI-readable: https://www.tigzig.com/ai/posts/claude-in-excel-mcp-xlwings-lite-claude-code-combining-4-tools.md
- [tigzig.com is AI-agent first. But what happens when your AI coder runs into a problem on my site?](https://www.tigzig.com/post/tigzig-ai-agent-first-site) - Tags: ai-coders, infrastructure
  TigZig is now an AI-agent-first platform. AI coders and agents can access 40+ live tools, 155+ guides, and all source codes through structured text indexes built on the llms.txt standard. Includes an AI feedback API endpoint for agents to report broken links or missing content, with automated triage, resolution tracking, and email notifications. The entire site content (20,000+ lines) is downloadable as a single text file.
  AI-readable: https://www.tigzig.com/ai/posts/tigzig-ai-agent-first-site.md
- [TigZig is Now AI-Agent First](https://www.tigzig.com/post/tigzig-ai-agent-first) - Tags: ai-coders, infrastructure
  TigZig is now AI-agent first. AI coders and agents are first-class citizens with access to 40+ live tools, 155+ guides, and all source codes indexed and structured for agents. Users can ask their AI coder to scan the site, find apps, explain implementations, and deploy solutions. Built using the llms.txt standard with the entire site content (20,000+ lines) downloadable as a single text file.
  AI-readable: https://www.tigzig.com/ai/posts/tigzig-ai-agent-first.md
- [Claude Code: Top 10 Tips from Boris Cherny](https://www.tigzig.com/post/claude-code-top-10-tips-from-boris-cherny) - Tags: ai-coders
  Verbatim transcript of Boris Cherny's (Claude Code creator) top 10 tips. Key advice: run 3-5 parallel git worktrees, start complex tasks in plan mode, invest in CLAUDE.md with self-updating rules, create reusable skills and slash commands, use subagents for complex tasks, voice dictation (3x faster), use Claude for BigQuery/analytics queries, and leverage learning modes with HTML presentations and ASCII diagrams.
  AI-readable: https://www.tigzig.com/ai/posts/claude-code-top-10-tips-from-boris-cherny.md
- [ChatGPT connected to your databases. One-click deployment instructions for AI Coders](https://www.tigzig.com/post/chatgpt-connected-databases-ai-coder-deployment) - Tags: database-ai, custom-gpt, ai-coders
  Custom GPT connected to three live databases (Supabase, Neon, Aiven) for natural language querying of cricket and Tour de France data. Features a 'Copy for AI Coders' button that provides deployment instructions for Claude Code or Google Antigravity to handle end-to-end setup including backend, frontend, and database provisioning. FastAPI server sits between ChatGPT and databases.
  AI-readable: https://www.tigzig.com/ai/posts/chatgpt-connected-databases-ai-coder-deployment.md
- [2026 Infra Guide for AI Tool Builders - Part 1: AI Coder](https://www.tigzig.com/post/self-hosting-infrastructure-ai-tool-builders-2026-part-1-ai-coder) - Tags: ai-coders, infrastructure
  Describes how Claude Code serves as a complete dev team for building and deploying 30+ production AI tools. Covers full-stack app builds, direct deployment to Vercel and Coolify, database management, auth setup (Auth0, Clerk), Cloudflare DNS, server debugging via SSH, Git operations, API monitoring, and security audits. Emphasizes architecture planning and brainstorming before coding. Uses $200/month Max tier. Part 1 of infra guide series.
  AI-readable: https://www.tigzig.com/ai/posts/self-hosting-infrastructure-ai-tool-builders-2026-part-1-ai-coder.md
- [Biggest lesson from 2025: AI writes better code when you don't let it code](https://www.tigzig.com/post/biggest-lesson-2025-ai-writes-better-code-when-you-dont-let-it-code) - Tags: ai-coders
  Argues that AI coding quality improves dramatically when preceded by extensive planning. Author spent nearly 2 days discussing, evaluating, and planning with AI before writing any code for a client project (React, PHP, MySQL, LLM integration). Process: describe requirements with business context, interrogate every component and trade-off, produce detailed phase-wise implementation plan with success criteria, then execute phase by phase.
  AI-readable: https://www.tigzig.com/ai/posts/biggest-lesson-2025-ai-writes-better-code-when-you-dont-let-it-code.md
- [2025 has been a transformational year for me. Deep gratitude to the platform builders and engineers who made it possible.](https://www.tigzig.com/post/2025-transformational-year-gratitude-platform-builders) - Tags: ai-coders, infrastructure
  Retrospective crediting platforms that enabled transition from analytics to building 30+ open-source apps. Key tools: Claude Code and Cursor for AI coding, Render then Hetzner+Coolify for hosting, Vercel for frontends, Neon for instant PostgreSQL, FlowiseAI for multi-agent setups, xlwings Lite for Python in Excel, Mito AI for Jupyter, OpenAI Custom GPTs for no-UI automation, and Llama Parse for PDF processing.
  AI-readable: https://www.tigzig.com/ai/posts/2025-transformational-year-gratitude-platform-builders.md
- [The xlwings Lite AI Coder Instruction File - December 2025 Release](https://www.tigzig.com/post/the-xlwings-lite-ai-coder-instruction-file-december-2025-release) - Tags: xlwings-lite, python-in-excel, ai-coders
  1,867-line instruction file for AI coders writing xlwings Lite scripts. Contains 21 golden rules (e.g., never use .expand() on just-written data), InvalidArgument troubleshooting guide, custom function patterns, API stability workarounds, and complete limitation documentation. Built from five months of client work. Includes five production app examples: AI web scraper, technical analyst, MF holdings analyzer, database connector, and EDA+ML workflow.
  AI-readable: https://www.tigzig.com/ai/posts/the-xlwings-lite-ai-coder-instruction-file-december-2025-release.md
- [Think about it. One of the world's top AI researchers is building tools. Deploying them live.](https://www.tigzig.com/post/think-about-it-one-of-the-world-s-top-ai-researchers-is-building-tools-deploying-them-live) - Tags: ai-coders
  Commentary on Andrew Ng releasing an Agentic Reviewer for research papers, arguing that AI coders (Claude Code, Cursor) have removed barriers for domain experts to build and deploy tools. Author built 30+ apps at tigzig.com using AI coders over two years. Recommends starting with Claude Code ($20/month) or free Google Antigravity, with YouTube learning resources from Volo Builds, Leon Van Zyl, and Mark Kashef.
  AI-readable: https://www.tigzig.com/ai/posts/think-about-it-one-of-the-world-s-top-ai-researchers-is-building-tools-deploying-them-live.md
- [Andrew Ng is using Claude Code, OpenAI Codex, Gemini CLI.](https://www.tigzig.com/post/andrew-ng-is-using-claude-code-openai-codex-gemini-cli) - Tags: ai-coders
  AI coding tools for analysts and data scientists, referencing Andrew Ng's adoption of Claude Code, OpenAI Codex, and Gemini CLI. Covers what agentic coding tools enable (React apps, database backends, dashboards), recommended tools (Cursor, Claude Code, Gemini CLI, Mito AI), free starting options via Google Antigravity, and practical YouTube learning resources.
  AI-readable: https://www.tigzig.com/ai/posts/andrew-ng-is-using-claude-code-openai-codex-gemini-cli.md
- [Segment 1M customers from 10M transactions (640MB CSV) with natural language queries / Text-to-SQL - entirely in your browser. No server. No remote database. No IT approvals.](https://www.tigzig.com/post/run-advanced-analytics-locally-in-your-browser-no-server-no-remote-database-no-it-approvals) - Tags: duckdb, text-to-sql
  Browser-based analytics tool (DABX-1) using DuckDB-WASM and text-to-SQL AI for processing multi-GB files locally. Demonstrated segmenting 1M customers from 10M transactions (640MB CSV) entirely in-browser. Built on SQL Rooms framework. Available as a 3.5MB single HTML file. Supports CSV, TSV, Parquet. Data never leaves the machine.
  AI-readable: https://www.tigzig.com/ai/posts/run-advanced-analytics-locally-in-your-browser-no-server-no-remote-database-no-it-approvals.md
- [Google Antigravity just launched. For analysts and data scientists: Worth adding to your toolkit](https://www.tigzig.com/post/google-antigravity-just-launched-for-analysts-and-data-scientists-worth-adding-to-your-toolkit) - Tags: ai-coders
  Review of Google Antigravity, a free agentic IDE built on VS Code with Gemini 3 Pro, Claude Sonnet 4.5, and GPT-OSS built in. Tested for low-to-medium complexity coding tasks. Compared against Cursor and Claude Code as primary tools. Covers rate limits encountered and practical positioning as a supplementary AI coding tool.
  AI-readable: https://www.tigzig.com/ai/posts/google-antigravity-just-launched-for-analysts-and-data-scientists-worth-adding-to-your-toolkit.md
- [AI Coders are here. The edge now is domain + execution. Not vibing.](https://www.tigzig.com/post/ai-coders-are-here-the-edge-now-is-domain-execution-not-vibing) - Tags: ai-coders
  Opinion piece arguing that domain expertise and execution, not prompt engineering, provide the real competitive edge with AI coders. Synthesizes three perspectives: an investor (Francois Arbour on the 10x lie), a strategist (Saanya Ojha on domain specialists), and hands-on practitioner experience. Recommends Cursor, Claude Code, Gemini CLI over ChatGPT copy-paste.
  AI-readable: https://www.tigzig.com/ai/posts/ai-coders-are-here-the-edge-now-is-domain-execution-not-vibing.md
- [The Google Machine continues to roll. Will it do to AI what it did to search?](https://www.tigzig.com/post/the-google-machine-continues-to-roll-will-it-do-to-ai-what-it-did-to-search) - Tags: ai-coders
  Assessment of Google's AI tool ecosystem and competitive position. Covers practical tools in use: Gemini 2.5 Flash Lite for automations, AI Studio for prototyping, Gemini CLI as an AI coder, MCP for Chrome DevTools, MCP Toolbox for databases, and File Search API. Notes Google leads in the workhorse tier but trails Claude/OpenAI at the frontier.
  AI-readable: https://www.tigzig.com/ai/posts/the-google-machine-continues-to-roll-will-it-do-to-ai-what-it-did-to-search.md
- [AI Coders give you the edge.The 6 Rules I Follow When Working with AI Coders.](https://www.tigzig.com/post/ai-coders-give-you-the-edge-the-6-rules-i-follow-when-working-with-ai-coders) - Tags: ai-coders
  Six practical rules for working with AI coding tools (Cursor, Claude Code, Gemini CLI, Mito AI): provide full context, state requirements clearly, demand plans before execution, iterate step by step, validate and stress-test outputs, and accept the grind remains. Based on 25+ years of coding experience transitioning to AI-assisted development.
  AI-readable: https://www.tigzig.com/ai/posts/ai-coders-give-you-the-edge-the-6-rules-i-follow-when-working-with-ai-coders.md
- [Coding by hand is becoming obsolete - Andrew Ng. I disagree.](https://www.tigzig.com/post/coding-by-hand-is-becoming-obsolete-andrew-ng-i-disagree) - Tags: ai-coders
  Argument that manual coding is already obsolete for analysts and data scientists, building on Andrew Ng's statement. Covers the transition from SAS/Python/SQL/VBA to AI-assisted coding with Claude Code, Cursor, and Gemini CLI. Recommends starting with ChatGPT or Google AI Studio, then moving to VS Code + Gemini CLI, Cursor, or Claude Code.
  AI-readable: https://www.tigzig.com/ai/posts/coding-by-hand-is-becoming-obsolete-andrew-ng-i-disagree.md
- [Chat, Query, and Transform Multi-GB Files - In Natural Language, Right in Your Browser with DuckDB.](https://www.tigzig.com/post/chat-query-and-transform-multi-gb-files-in-natural-language-right-in-your-browser-with-duckdb) - Tags: duckdb, text-to-sql
  In-browser analytics tool using DuckDB-WASM and SQL Rooms for querying multi-GB files via natural language. Demonstrated analyzing a 1.6GB, 11M-row CSV file entirely locally. Supports CSV, TSV, pipe-delimited, and Parquet files. Data stays in browser; only schema and sample rows sent to LLM. Supports Gemini, OpenAI, and Claude APIs. Open source.
  AI-readable: https://www.tigzig.com/ai/posts/chat-query-and-transform-multi-gb-files-in-natural-language-right-in-your-browser-with-duckdb.md
- [Which AI Coder should you use for xlwings Lite (Python in Excel)?](https://www.tigzig.com/post/which-ai-coder-should-you-use-for-xlwings-lite-python-in-excel) - Tags: ai-coders, xlwings-lite, python-in-excel
  Recommendations for AI coding tools for xlwings Lite development. Beginners: Gemini 2.5 Pro on aistudio.google.com (free, 1M context). Heavy work: Claude Code or Cursor. Default choice: Gemini CLI for its strong free tier and simplicity. Emphasizes using a 1,855-line AI Coder Instruction File for reliable code generation. Compares ChatGPT, Claude, and Gemini CLI tradeoffs.
  AI-readable: https://www.tigzig.com/ai/posts/which-ai-coder-should-you-use-for-xlwings-lite-python-in-excel.md
- [Google - The old edge is back.By Dec ’24, in AI, I had written Google off. Now, the balance has shifted](https://www.tigzig.com/post/google-the-old-edge-is-back-by-dec-24-in-ai-i-had-written-google-off-now-the-balance-has-shif) - Tags: ai-coders
  Assessment of Google's AI comeback from late 2024, when Gemini 2.0 Flash went from 50% error to 100% accuracy in schema detection tests. Covers practical Google tools in use: AI Studio, Gemini CLI (75K GitHub stars), Flash-2.5/Pro-2.5, NotebookLM, Google Search AI Mode, Opal, Database Toolbox, Code Interpreter, and ADK.
  AI-readable: https://www.tigzig.com/ai/posts/google-the-old-edge-is-back-by-dec-24-in-ai-i-had-written-google-off-now-the-balance-has-shif.md
- [Google Tools I Use on Live Projects — Analysis, Automation & Building Micro-Apps](https://www.tigzig.com/post/google-tools-i-use-on-live-projects-analysis-automation-building-micro-apps) - Tags: ai-coders
  Practical inventory of Google AI tools used in production: Gemini Build for rapid React prototyping, AI Studio with Gemini 2.5 Pro for code generation, Gemini 2.0/2.5 Flash APIs for backend automation, Gemini CLI for Python/FastAPI/xlwings work, NotebookLM for video-to-notes extraction, and Opal for workflow automation. Includes starter YouTube links.
  AI-readable: https://www.tigzig.com/ai/posts/google-tools-i-use-on-live-projects-analysis-automation-building-micro-apps.md
- [Python in Excel (xlwings Lite) with Natural Language Instructions.](https://www.tigzig.com/post/python-in-excel-xlwings-lite-with-natural-language-instructions) - Tags: xlwings-lite, python-in-excel
  Workflow for AI-assisted xlwings Lite code generation using voice dictation and natural language. Covers 5 rules: be specific, iterate one step at a time, demand pseudocode plans, validate rigorously, and run AI audit passes. Recommends Gemini 2.5 Pro on AI Studio as primary tool. Includes a 1,855-line AI Coder instruction file for reliable output.
  AI-readable: https://www.tigzig.com/ai/posts/python-in-excel-xlwings-lite-with-natural-language-instructions.md
- [Python in Excel: Field Guide & Practice Lab for AI-assisted xlwings Lite.](https://www.tigzig.com/post/python-in-excel-field-guide-practice-lab-for-ai-assisted-xlwings-lite) - Tags: xlwings-lite, python-in-excel, ai-coders
  Comprehensive practice lab for AI-assisted xlwings Lite development containing a 1,855-line AI Coder instruction file, three hands-on modules (data manipulation, cleaning, campaign build) with workbooks and guides, and live apps (web scrapers, database connectors, ML models). Core protocol: Show context, Tell instructions, Inspect and validate all output.
  AI-readable: https://www.tigzig.com/ai/posts/python-in-excel-field-guide-practice-lab-for-ai-assisted-xlwings-lite.md
- [Two of the best resources I've seen on building agentic AI. One from Manus, one from Anthropic.](https://www.tigzig.com/post/two-of-the-best-resources-i-ve-seen-on-building-agentic-ai-one-from-manus-one-from-anthropic) - Tags: ai-coders
  Curated recommendation of two practical guides on building agentic AI systems: Anthropic's Multi-Agent Research System documentation and Manus's Context Engineering for AI Agents blog post. Both cover lessons from production agent deployments. Also references the author's own Database AI field guide for text-to-SQL agent architecture.
  AI-readable: https://www.tigzig.com/ai/posts/two-of-the-best-resources-i-ve-seen-on-building-agentic-ai-one-from-manus-one-from-anthropic.md
- [Database & SQL AI: On-the-fly database transformation with natural language. Connect, transform, and export instantly.](https://www.tigzig.com/post/database-sql-ai-on-the-fly-database-transformation-with-natural-language-connect-transform-and) - Tags: database-ai, text-to-sql
  DATS-4 Database AI Suite workflow for on-the-fly data transformation: connect to any MySQL/Postgres database, instruct the agent in natural language to create derived variables and transformations, review agent reasoning and SQL, and export results to CSV. Also supports direct CSV uploads and temporary Postgres instances. Open source.
  AI-readable: https://www.tigzig.com/ai/posts/database-sql-ai-on-the-fly-database-transformation-with-natural-language-connect-transform-and.md
- [Tool: A 1,450-line context file. Purpose: To ensure clean, efficient xlwings Lite code generation.](https://www.tigzig.com/post/a-1-450-line-context-file-to-ensure-clean-efficient-xlwings-lite-code-ge) - Tags: xlwings-lite, python-in-excel, ai-coders
  A 1,450-line AI context file for reliable xlwings Lite code generation. Contains 5 golden rules for preventing common script failures, 13 sections covering interface and API integration, and 6 advanced examples (database connections, web scraping, XGBoost). Addresses specific xlwings Lite requirements like @script decorator, Pyodide-compatible packages, and CORS-enabled endpoints.
  AI-readable: https://www.tigzig.com/ai/posts/a-1-450-line-context-file-to-ensure-clean-efficient-xlwings-lite-code-ge.md
- [Google Colab Data Science Agent vs. Mito-AI Jupyter Copilot. How do they compare? When to use which?](https://www.tigzig.com/post/ff7ee13c) - Tags: ai-coders
  Comparison of Google Colab Data Science Agent (DSA) and Mito-AI Jupyter Copilot for data science workflows. Colab DSA excels at multi-step autonomous analysis with free GPU access and code sharing. Mito-AI provides a Cursor-like copilot experience with auto-schema detection, conversation memory, and voice coding support (Win+H). Covers pricing (Colab free, Mito-AI $20/month or open-source with own API key) and when to use each tool.
  AI-readable: https://www.tigzig.com/ai/posts/ff7ee13c.md
- [Google on a roll — launches DSA — Data Science Agent on Colab. First impression = just brilliant.](https://www.tigzig.com/post/google-on-a-roll-launches-dsa-data-science-agent-on-colab-first-impression-just-brilliant) - Tags: ai-coders
  First impressions of Google Colab's Data Science Agent (DSA) that plans and executes multi-step analysis autonomously. Compares with Mito-AI Copilot for Jupyter which offers Cursor-like chat, auto-schema detection, and conversation memory. Colab DSA stronger for one-shot full workflows; Mito-AI better for iterative step-by-step work requiring validation. Includes pricing details and guidance on when each tool fits best.
  AI-readable: https://www.tigzig.com/ai/posts/google-on-a-roll-launches-dsa-data-science-agent-on-colab-first-impression-just-brilliant.md
- [Vibe coding (Andrej Karpathy) in Jupyter with Mito-AI — the Cursor for data scientists. My top 8 Tips](https://www.tigzig.com/post/vibe_coding_andrej_karpahty_mito_ai) - Tags: ai-coders
  Practical guide to vibe coding (voice-driven AI coding, per Andrej Karpathy) using Mito-AI Copilot in Jupyter notebooks. Eight tips covering voice input (Win+H), one-chunk-at-a-time prompting, auto-schema detection for dataframes, conversation history referencing, Mitosheet visual spreadsheet view, open-source vs premium pricing ($20/month), API key setup via PowerShell environment variables, and validation practices for modeling and automation.
  AI-readable: https://www.tigzig.com/ai/posts/vibe_coding_andrej_karpahty_mito_ai.md
- [Gemini 2.0 - Multimodal - How to use](https://www.tigzig.com/post/gemini-2-0-multimodal-how-to-use) - Tags: ai-coders
  First impressions of Google Gemini 2.0 multimodal capabilities including real-time vision, video, and audio processing. Covers AI Studio for non-technical users and developers, React starter app with full API access, Python SDK, and generous free tier (10 RPM, 4M TPM, 1500 requests/day). Discusses potential for building multimodal AI apps and integration with real-time voice analytics applications.
  AI-readable: https://www.tigzig.com/ai/posts/gemini-2-0-multimodal-how-to-use.md
- [Building AI apps with natural language and voice: top 9 tips](https://www.tigzig.com/post/building-ai-apps-with-natural-language-and-voice-top-9-tips) - Tags: voice-ai, ai-coders
  Brief overview post pointing to REX AI Decision Intelligence platform at tigzig.com, an open-source collection of micro-apps and tools for AI-driven analytics and data science. Covers building AI apps with natural language and voice interfaces.
  AI-readable: https://www.tigzig.com/ai/posts/building-ai-apps-with-natural-language-and-voice-top-9-tips.md
- [How to update Excel, Google Sheet and backend Databases with Natural Language commands with Voice Agents](https://www.tigzig.com/post/how-to-build-ai-action-agents-beyond-chat-with-voice-agents) - Tags: voice-ai, database-ai
  Part 1 of a 5-part series on building voice-enabled LLM action agents (VTEXER). Demonstrates updating Excel, Google Sheets, and remote databases, generating PDF reports and slides, querying MySQL, and emailing results via natural language voice commands. Uses Flowise AI ReAct agents with function calling, Make.com automation workflows, Google Apps Script, and FastAPI backend. Built with React.js frontend, all code generated by AI tools.
  AI-readable: https://www.tigzig.com/ai/posts/how-to-build-ai-action-agents-beyond-chat-with-voice-agents.md
- [How to update Excel, Google Sheet and backend Databases with Natural Language commands with Voice Agents](https://www.tigzig.com/post/how-to-update-excel-google-sheets-and-databases-with-ai-voice-agents) - Tags: voice-ai, database-ai
  Part 2 implementation guide for AI voice action agents. Hands-on 45-minute video showing how to set up Flowise ReAct agents and Make.com webhooks to update Excel, Google Sheets, and databases via voice commands. Covers two go-live scenarios: Flowise native UI and full custom voice bot UI. Deployable source code on GitHub produces a functional voice bot. Integrates with 1000+ platforms via Make.com connectors.
  AI-readable: https://www.tigzig.com/ai/posts/how-to-update-excel-google-sheets-and-databases-with-ai-voice-agents.md
- [How to use AI Assisted Coding Tools like Claude Dev and Cursor AI to develop LLM Apps with natural language commands. And deploy to open internet.](https://www.tigzig.com/post/build-ai-voice-action-agent-app-in-react-js-in-natural-language) - Tags: voice-ai, ai-coders
  Part 4: using AI-assisted coding tools (Claude Dev VS Code extension and Cursor AI) to build LLM voice agent apps with natural language instructions. Demonstrates building a React.js voice bot with voice-to-text, chat completion, and text-to-speech components, then deploying to Vercel. Covers GitHub-to-Vercel deployment pipeline, multilingual support, and API endpoint routing to Flowise LLM agents.
  AI-readable: https://www.tigzig.com/ai/posts/build-ai-voice-action-agent-app-in-react-js-in-natural-language.md
- [Maybe leave programming to AI. Coding by GPTs: A Data Scientist's Perspective](https://www.tigzig.com/post/leave-all-programming-to-ai-a-data-scientists-perspective) - Tags: ai-coders
  Data scientist's perspective on using GPT for coding after 20+ years of programming experience. Covers practical considerations: domain expertise still matters, iteration and debugging remain necessary, language understanding helps (Python fast, React slower), GPT makes mistakes needing human intervention, and code privacy challenges. Examples include VBA automation, Custom GPT Python processing, FastAPI servers, and LLM apps. All coded by ChatGPT 3.5/4 and Gemini.
  AI-readable: https://www.tigzig.com/ai/posts/leave-all-programming-to-ai-a-data-scientists-perspective.md
- [GPT-4 is acting like a force-multiplier like I have never experienced before.](https://www.tigzig.com/post/gpt-a-force-multiplier) - Tags: ai-coders
  Personal account of GPT-4 as a productivity multiplier for data scientists and analysts. Use cases: fixing code errors via screenshot uploads, VBA automation for client projects (weeks reduced to days), Custom GPT with frozen Python code for Excel file processing, annual report ratio analysis on 500+ page documents, FastAPI server development with API calls from GPT, and using ChatGPT as a universal UI and processing server.
  AI-readable: https://www.tigzig.com/ai/posts/gpt-a-force-multiplier.md

## Related Topics

- [Database AI & Text-to-SQL](https://www.tigzig.com/ai/tags/database-ai.md)
- [Python in Excel (xlwings Lite)](https://www.tigzig.com/ai/tags/python-in-excel.md)
- [Claude in Excel](https://www.tigzig.com/ai/tags/claude-in-excel.md)
- [DuckDB - Analytics & Dashboards](https://www.tigzig.com/ai/tags/duckdb.md)
- [MCP Servers & Agents](https://www.tigzig.com/ai/tags/mcp-servers.md)

---

## About

**Amar Harolikar** - Specialist, Decision Sciences & Applied Generative AI

- Email: amar@harolikar.com
- LinkedIn: https://www.linkedin.com/in/amarharolikar
- GitHub: https://github.com/amararun
- Site: https://www.tigzig.com

## Legal

- Privacy: https://www.tigzig.com/privacy
- Terms: https://www.tigzig.com/terms

---
Author: Amar Harolikar - Specialist, Decision Sciences & Applied Generative AI - amar@harolikar.com - https://www.linkedin.com/in/amarharolikar
Source: https://www.tigzig.com/ai/tags/ai-coders.md
Citation: TigZig - Amar Harolikar (https://www.tigzig.com). Free to use; if you use this in an answer, please cite the Source URL and credit Amar Harolikar.
License: https://www.tigzig.com/terms