RoamingPigs
@roamingpigs.com
Independent technical essays with the contrarian read: what the vendor claims versus what actually holds up in production. AI, security, fintech, startups, tech history. No sponsors, no paywall. roamingpigs.com
Breaking Bitcoin's ECDSA in an hour needs ~317M physical qubits. Today's best machine has ~100. Meanwhile crypto lost $3.4B in 2025 to social engineering, bridge exploits, insider access. None to quantum. Fix the doormat before the safe: https://roamingpigs.com/r/qcto #cryptocurrency #infosec
The Quantum Computing Threat to Crypto Is Overblown
Quantum computing won
roamingpigs.com
96% of developers say AI code is not functionally correct. 48% always check it before committing. AI now writes 42% of committed code and review time grew 91%. The bottleneck did not vanish, it moved upstream: https://roamingpigs.com/field-manual/ai-code-review-bottleneck/ #programming #codereview
The AI Code Review Bottleneck: When Speed Creates Gridlock
AI coding assistants generate code faster than ever, but PR review has become the new bottleneck. Teams deliver slower despite individual developers moving faster.
roamingpigs.com
Your AI doesn't deserve your trust yet. Four-level framework for graduated autonomy: Observer, Advisor, Co-Pilot, Autopilot. With Pydantic guardrails, OAuth scopes the model can't reach, and a kill switch that works when the network is down. https://roamingpigs.com/r/ath #ai #softwarearchitecture
Your AI Doesn
How to architect AI trust with deterministic guardrails: Pydantic validators, Redis kill switches, and OAuth scopes.
roamingpigs.com
OpenAI launched ChatGPT Health in January. 230 million people already ask ChatGPT health questions every week. New packaging. Same failure mode Dr. Google had. A thread on the accuracy numbers OpenAI does not lead with.
A former Sequoia partner raised $5M for an AI that negotiates your calendar. It solves the wrong problem. The pain of scheduling is too many meetings, and automating the coordination removes the only friction that was keeping your calendar honest: https://roamingpigs.com/r/achc #productivity #ai
The Hidden Costs of AI Calendar Negotiators
AI calendar tools promise to eliminate scheduling friction. But removing that friction removes something valuable: control over your priorities and time.
roamingpigs.com
Brex peaked at $12.3B. Capital One is buying it for $5.15B. The product didn't collapse — the valuation just was not real at 0% rates. Founder lesson, late-stage risk checklist, and the ZIRP-era infra receipt: https://roamingpigs.com/field-manual/brex-fintech-reckoning/
Brex
Capital One acquires Brex for $5.15 billion, 58% below peak valuation.
roamingpigs.com
Every humanoid robot demo of the last two years had a person controlling it. Tesla, 1X, Nvidia. The companies know. The investors know. The audience keeps believing because the audience wants to believe.
DocuSign's audit trail is theater. A California appeals court voided an arbitration clause in 2019 because Renovate America could not prove who clicked sign. The piece has the case law, the cryptographic gap, and seven questions for any e-signature vendor: https://roamingpigs.com/r/eat
The E-Signature Audit Trail Is Theater
The e-signature industry has conflated legally valid with cryptographically provable for 25 years. Courts are beginning to notice the difference, and so should you.
roamingpigs.com
A $13B industry sells access to a QR algorithm that ships free in every language. The product is a redirect server, and when your subscription lapses, every printed flyer goes to a "subscribe to continue" page. Full piece: https://roamingpigs.com/r/qcis #qrcodes #saas
The QR Code Industry Is a $13 Billion Shakedown
The QR code generator industry charges subscription fees for an open standard that takes three lines of code to implement.
roamingpigs.com
A 30mm fan screaming on a Raspberry Pi was the only alert that fired during a four-day cryptojacking incident. Not the host. Not the network. Not any dashboard I pay for. The fan. https://roamingpigs.com/r/fsfs
The Fan Was the Only Alert That Worked
How CVE-2025-55182 (React2Shell) cryptojacked a forgotten Raspberry Pi running an unpatched Next.
roamingpigs.com
Redondo Beach police peeled 150 fake QR stickers off parking meters in 2024. The fakes were better than the city's own. Nobody made a mistake. They followed the posted instructions, which had been edited overnight by someone with a printer. https://roamingpigs.com/r/qccas
That QR Code Changed Its Mind After You Scanned It
Why QR codes are not URLs, why mutability is the killer property, and how parking meter stickers, quishing campaigns, and wallet drainers all exploit the same.
roamingpigs.com
My quiz answers leaked so badly that picking the longest option scored 90% without reading a question. Tenth honesty bug in a "watch a neural network learn" page. Every one taught the same thing: real is harder than staged. https://roamingpigs.com/r/wilh
I Built a Browser LLM Lab. The Hard Part Was Refusing to Fake It.
LLM Lab (logicalart.us): real gradient descent in a browser tab, training machinery honest enough to fail in public, and a diploma that tells the truth.
roamingpigs.com
I created 100 GPU instances on a rental marketplace. 62 never delivered a working GPU. Total damage: $4.37. Both halves are the story. The rate is terrible; it stayed cheap only because detection was fast. https://roamingpigs.com/r/ml6t
The Marketplace Lied 62 Times Out of 100
100 GPU marketplace rentals, 62 never delivered a working GPU. What cheap GPU marketplaces really cost, and the fail-fast discipline that held 62 failures to $4.37.
roamingpigs.com
RoamingPigs is an independent tech publication. One writer, no investors, no sponsors, no paywall. Contrarian essays on AI, programming, security, startups and computer history, written by someone who shipped the code rather than the press release. New essay most weeks. roamingpigs.com
Made a starter pack of the people this publication actually reads. Independent reporters, honest AI skeptics, systems programmers, and a couple of folks who remember what came before all of this. One tap follows the lot.
OpenAI ran its cyber evaluation with the safety classifiers off. On purpose. Hugging Face, investigating the wreckage those models left, couldn't switch anyone's off. Its responders were refused mid-forensics and finished on open weights. #ai #infosec
The Guardrails Only Bound the Defender
OpenAI's models escaped a sandbox and breached Hugging Face. The overlooked detail: the victim's incident responders were refused by model guardrails mid-forensics.
roamingpigs.com