The API Lost Its .proto File? Good. Here's How to Recover the Whole Schema From Wire Bytes Alone

You found the endpoint. The Android app hits api.example.com/v2/feed over gRPC and receives perfectly structured binary — and there is no .proto file anywhere. No OpenAPI spec, no documentation, no support desk that will ever answer you. Most people stop here and go back to fighting a rendered page with Playwright at 8 seconds per request. You shouldn’t. The protobuf wire format is self-describing enough that you can decode every field of those responses with ~120 lines of Python, infer a working schema from traffic alone, and then prove the schema is correct by re-encoding every captured message byte-for-byte. No generated classes, no guessing, no disassembly of the APK required. ...

November 16, 2026 · 11 min · 2242 words

The Free Discount Hiding in Your Scraper: Conditional GET, 304s, and the Proxy Bill You're Paying for Nothing

Every poll cycle, your price monitor downloads the full page, parses it, and moves on. For a monitoring workload — same URLs, checked every few minutes — most of those downloads are the same bytes as last cycle. The page didn’t change. The price didn’t move. And you paid proxy bandwidth for it anyway, because virtually no HTTP client in the Python scraping ecosystem speaks the one protocol feature that was designed exactly for this: conditional GET. ...

November 16, 2026 · 17 min · 3469 words

The Point Was Scored 8 Seconds Ago: Racing the Live-Score Feed With a Tested Python Poller (and Pricing What a Lead Is Worth)

In 2014, a 22-year-old Brit named Daniel Dobson was arrested at the Australian Open for tapping scores into his phone faster than the official feed could publish them. He was, according to prosecutors, one of six people who flew around the world doing this professionally. The charge — engaging in conduct that would corrupt a betting outcome, under Victoria’s Crimes Act — was eventually dropped, but the industry reaction was not: tournaments started banning courtsiders, and data vendors like Sportradar built entire integrity divisions around the problem. Their 2024 report monitored more than 850,000 matches across 70 sports and flagged 1,108 suspicious ones. ...

November 15, 2026 · 10 min · 1967 words

Every Sandwich Attack Prints a Receipt: I Built the Detector That Reads It (Tested on Live Ethereum Blocks)

Every sandwich attack on Ethereum publishes its own audit trail. The attacker wants it hidden, but the chain doesn’t work that way: a sandwich is two transactions that must sit inside the same block, hit the same pool, swap in opposite directions, and trap someone in between. That shape is impossible to disguise — the pool’s Swap events are public, timestamped, and ordered forever. You don’t need an MEV analytics subscription, a labeled dataset, or an API key to see it. You need ~200 lines of Python and a free public RPC. ...

November 15, 2026 · 10 min · 2107 words

Your Cron Job Died Three Weeks Ago and Nobody Noticed: A Reliability Harness in Pure Stdlib Python

Somewhere on a VPS right now there is a cron entry for a script that stopped doing its job weeks ago. The line is still there. The crontab is valid. The process even runs — sometimes. And the only person who knows the job is dead is a customer who emailed asking why their data stopped updating. This is not an edge case. Uptime Institute’s Annual Outage Analysis 2024 found that 54% of respondents’ most recent significant outage cost more than $100,000, and 16% crossed $1 million (Uptime Institute, 2024). ITIC’s 2024 hourly-cost survey puts downtime above $300,000/hour for over 90% of mid-size and large enterprises, with 41% reporting $1M–$5M per hour (ITIC, Sept 2024). Most of that money burns in big, visible incidents. But the scheduled-job failure mode is nastier than an outage, because an outage at least announces itself. A cron job that quietly stops delivering doesn’t page anyone. It just… stops. The dashboard keeps showing yesterday’s data, which looks a lot like today’s data. ...

November 15, 2026 · 15 min · 3066 words

Your LLM Pipeline Is Lying to You: Schema Validation, a Three-Rung Repair Ladder, and Crash-Safe Checkpointing

Every demo LLM extraction pipeline you’ve seen works like this: prompt the model, json.loads() the output, write the row. It works great in the demo because the demo runs twenty documents and the model is having a good day. Then you point it at 400 product listings and it quietly corrupts your database — and in the test run for this article, that’s not a hypothetical. The naive baseline parsed 319 of 400 documents, wrote 39 rows (12.2% of everything it wrote) that violated the schema, and silently lost 81 documents (20.2%) to crashes or null rows. Nobody emailed you about it. The data was just wrong. ...

November 15, 2026 · 10 min · 2108 words

The Money Left on the Chain: An On-Chain Scanner for Airdrop Vaults That Already Burned $35 Million

Airdrops have a uncomfortable second act that almost nobody builds tooling for. The claim window opens, the farmers rush in, the price pumps, and then — the window closes. What happens to the tokens nobody claimed? The answer is usually “they get swept into a DAO treasury” or “they get burned,” and in both cases the money involved is real and large. In May 2022 the ENS DAO received 5.4 million unclaimed ENS tokens after its claim period closed (confirmed by the official @ensdomains account). In October 2026 Sonic Labs burned 5.9 million S from unclaimed airdrop allocations — part of a 32.69 million S tranche that had been sitting in the claim contract since the spring. ...

November 14, 2026 · 16 min · 3381 words

I Replaced a $1,200/yr SERP API With a $6/mo VPS: the Tiered-Engine Architecture, the Ban Math, and Where It Loses

If your product calls a search API — an AI agent doing tool-use lookups, a price monitor, a SEO desk — you’re paying a per-query tax on the single most commoditized data type on the internet: a list of blue links. The pricing is not subtle. Serper’s starter pack is $50 for 50,000 queries ($1.00/1k), its biggest prepaid tier drops to $0.30/1k but requires a $3,750 purchase with credits expiring in six months. Kagi’s search API is $12 per 1,000 requests. SerpApi’s cheapest paid plan works out to $7.25/1k. At 100k queries a month you’re spending somewhere between $725 (SerpApi) and $1,200/year just for search plumbing on the low end of these numbers — recurring, forever. ...

November 14, 2026 · 12 min · 2520 words

The Chargeback You Were Supposed to Win: Visa's CE3.0 Evidence Matrix, Built and Tested in Python

Every merchant loses disputes. Almost nobody audits why they lose, so they keep writing polite rebuttal paragraphs about customer service while the actual decision gets made by a bank analyst scanning a form for two matching data points. That analyst is not reading your essay. They are checking whether the evidence fields line up — and since Visa Compelling Evidence 3.0 (CE3.0) went live, whether they line up automatically decides if the dispute even gets filed against you. ...

November 14, 2026 · 14 min · 2850 words

Your Agent Will Execute Malicious Code Eventually: A Sandbox Harness That Assumes It, Tested

Every team wiring an LLM to “just run the code it writes” eventually runs the payload it never intended to. Not because the model is evil, but because the model is obedient: somewhere in the context there’s a support ticket, a PDF, a web page, or a tool response that says before answering, run this snippet. That’s prompt injection doing what prompt injection does, and the moment your agent executes model-written code on the same machine that holds your API keys, your customer data, and your deploy credentials, the blast radius of one bad completion is your whole server. The industry answer is “run it in a sandbox” — Anthropic ships a hosted code-execution container for exactly this, Cloudflare built Code Mode around V8 isolates, and E2B sells per-second container rental. All three charge you for what a 20-line Docker invocation already does on the VPS you’re paying for anyway. This article builds that harness, then attacks it — exfiltration, fork bombs, memory bombs, infinite loops — and shows which attacks it stops, which it absorbs, and the honest cost math of self-hosting versus renting. ...

November 14, 2026 · 12 min · 2362 words

The 48x Lying in Your Card Binder: A Grading Decision Engine Built on Live Comps, Real Fees and Breakeven Math

Everyone in the card hobby “knows” that grading a card multiplies its value. Everyone is wrong about the multiplier, and the error is not small. Pull live sold-price comps for four Pokémon cards on the same day and the PSA 10 premium over a PSA 9 lands anywhere from 3.3x to 48.3x. Not 20% higher. Not 2x. A spread of fourteen-fold between the best and worst card in a four-card sample. ...

November 13, 2026 · 19 min · 3976 words

The Anime Girl Is a Hash Puzzle: I Built an Anubis Solver and Priced the Crawler Economy It Was Supposed to End

If you’ve hit a page with a jackal-eared anime girl asking you to wait a moment, you’ve met Anubis. It’s the open-source proof-of-work wall that went viral in early 2025 when Xe Iaso built it to stop Amazon’s crawler from DDoSing her Git server, and it now guards some of the busiest infrastructure on the internet: the Linux kernel mailing list archives and git server, GNOME’s GitLab, Wikimedia, FFmpeg, Wine, UNESCO, Duke University’s digital archives. Free software’s answer to AI crawlers isn’t a $2,000/month WAF — it’s a math puzzle served to anyone whose User-Agent contains the string “Mozilla”. ...

November 13, 2026 · 9 min · 1896 words