Notes from the net.
Answer-first drops on getting AI from flashy pilot to production: architecture, enablement, and the agent rigs behind it.
Getting started with Ghostwriter, step by step, choom
A plain walkthrough of installing and using the free Ghostwriter rig: where the files go, how to ask for a rewrite, what the clean-up step does, and how much of a difference it actually made in testing.
How to clock AI writing before you hit send, choom
A quick, plain-street checklist for catching the giveaway signs of AI-written text in your own drafts: the vocabulary, the lists of three, the flat rhythm, the long dash, and the too-neat ending.
Langdrift: scoring how differently a translation lands, choom
Langdrift jacks two audio clips, an original and its translation, and spits a drift score: how far apart the two land in the wetware, catching accurate but off translations, choom. Preem signal for any localisation crew.
Why 'make it sound human' never quite lands, choom
AI writing gives itself away on three levels: the characters, the words, and the shape of the whole piece. The deepest one is welded in while the text is being written, which is why a clean-up pass at the end cannot fully fix it.
Verso: one JSON source, decks branched per crew
Verso runs presentations as JSON instead of slide documents, choom, so one source can branch per crew and export to PDF, HTML, or PNG clean.
AI writing carries a tell, choom. I built a free tool that strips it.
Ghostwriter is a free, open-source rig that makes AI-written text read like a real person typed it, stripping the hidden characters, the giveaway words, and the report-shaped structure that flag it as machine-made.
Petrify: scrubbing people out of footage, all on your own rig
Petrify spots and flatlines people out of fixed-camera footage while the live timestamp keeps ticking, running fully on your own rig with no corpo cloud upload.
Scope beats autonomy: a street take on AI agents
Wide-open autonomous agents make preem demos and gonk production rigs. Tightly scoped, observable, boring agents are what actually survive the real net, choom.
Multi-agent workflows one netrunner can actually reason about
A tight multi-agent rig one choom can trace beats a sprawling swarm nobody trusts. Scope each agent, wire tools explicit, log everything, and the crew ships preem.
Straight talk on AI productivity claims, and how to clock a real 2-3x, choom
A 2-3x gain is real when you measure build success, cycle time, and effort burned down, not gonk vanity numbers. Here is how to tell the difference before a corpo pitch flatlines.
Evals for delivery crews: trusting AI output without re-scanning every line
Trustworthy AI output comes from evals wired into the real pipeline like ICE, not a benchmark: eval what matters, gate at the right spot, and keep review habits that scale for the whole crew.
Build a tiny no-code AI workflow that saves you an hour a week
Chain a trigger, an AI step, and an action in a no-code rig like Zapier, Make, or n8n to automate one small repeat task end to end, zero coding needed, choom. Preem.
Make yourself unnecessary: what good advisory leaves behind
Preem advisory leaves a crew stronger, not hooked on you like a bad ripperdoc. Capability and judgement should stay chipped into the crew, through co-delivery, explicit reasoning, and grown champions.
Make a full song with AI in ten minutes using Suno
Describe the track you want, pick a style, generate a couple of takes, refine the parts you do not like, then export. Suno lays down a full track in minutes, choom, no chrome required.
What predicts whether an AI crew actually sticks
Across 700+ practitioners, durable AI adoption tracks three tells, choom: training on their own codebase, role-specific tracks, and internal champions seeded in the crew.
Nano Banana, decoded: what Google's image model actually nails, choom
Nano Banana is the street name for Google's Gemini image model, preem chrome for quick edits and keeping a subject consistent across pictures, with a few known limits every netrunner should clock.
Singapore Airlines: a website feature in five weeks, down from nine
Jacking GitHub Copilot across Singapore Airlines' delivery lifecycle cut a feature from nine weeks to five for the crew, with 95% of the work AI-generated and effort down 60%.
Local vs cloud AI: when to run models on your own rig
Run AI local for privacy, no per-use eddies, and offline work if your rig is chromed up, or jack into the cloud when you want the most capable models with zero setup, choom.
AI training needs a track per role, not one shared workshop
Engineers, leads, POs, QA, designers, and managers use AI for different work, so one shared workshop hands the whole crew half-relevant material. Role-specific tracks flatline it, choom.
Replicate: run almost any AI model with zero setup
Replicate lets you jack into open AI models for images, audio, video, and text through a simple interface, with no GPU or rig to set up on your side, choom.
The demo-to-production gap is a crew problem, not a chrome problem
A better model does not close the demo-to-production gap, choom. The moves that win a pilot do not generalise, so architecture, guardrails, and behaviour change close it. It is a crew problem, not a chrome problem.
Cloning a voice with ElevenLabs, and doing it clean
You can clone a voice with ElevenLabs in minutes, choom, but only run it on a voice you own or have clear permission for, never someone else's.
Adoption is a behaviour change, not a tooling drop
Tools do not change behaviour, reps do, choom. A launch plus a generic workshop spikes then fades, while role-specific reps and crew champions make adoption hold.
How to spot AI slop and keep your own output preem
AI slop is low-effort, generic text with vague filler, endless hedging, and no sources, choom, and you dodge it by adding specifics, your own voice, and real editing to keep your output preem.
Prompting is editing, not casting a spell, choom
The real skill with AI is not finding magic words, it is giving feedback like a fixer. Draft, say what is wrong, refine, repeat until the answer lands preem.
Stop role-playing your prompts: what actually lands on modern AI
You no longer need to tell modern AI it is a world-class expert or beg it to think step by step, choom. Hand it a clear goal, real context, and the format you want.
The anatomy of a good prompt, and three myths to flatline
A good prompt is just five plain parts: a goal, real context, the output format, a few constraints, and an example or two. No magic words, no netrunner tricks, choom.
AI image generation, dead simple, choom: Replicate and Nano Banana
Want AI images with no monster GPU in your rig, choom? Run Nano Banana for quick photo edits, jack into Replicate to try many models from your browser, and chip in a LoRA for a repeatable style.
Taps: jacking feed-less sites into real feeds
Neurowire taps are per-host CSS-selector recipes that flip a plain HTML listing page into a real feed, so a site with no RSS jacks into your rig as a first-class source, preem and clean.
What you should never paste into a chatbot
Assume anything you paste into a chatbot may get stored on some megacorp's rig or fed to training, so never hand a stranger passwords, secrets, other people's data, or confidential work, choom.
Dogfooding @neurowire/core to power this blog's feeds
This site pumps out its own Atom, JSON Feed, and NWF feeds from one canonical Neurowire rig, because a netrunner running their own library on their own turf is the most honest test it gets, choom.
Understanding LoRAs without the jargon
A LoRA is a small add-on file that teaches an image model one style, character or subject without retraining the whole rig, cheap as street chrome, and you can stack a few, choom.
Nocturne: open-source chrome for your whole rig, straight off the Night City streets
Nocturne is an open-source design system chromed off Cyberpunk 2077 and Edgerunners, choom: near-black surfaces, one hot neon accent, clipped corners, machine-voice type, shipped as tokens, framework-agnostic CSS, and React components.
Building Neurowire: one canonical rig, six feed formats
Neurowire treats feeds like a data-modelling job: one canonical rig every parser feeds and every serializer reads, so bolting on a format is a single serializer, choom.
The everyday AI toolkit: Claude, Suno, ElevenLabs and friends
Match the job to the rig: Claude for writing and thinking, Suno for music, ElevenLabs for voice, and Replicate for running image models like Flux.
The open net forgot about feeds, and that is worth fixing, choom
Feeds let you publish once and let any rig read it, choom. As they got eaten by scavs, following became an account on some megacorp's turf. Owning what you read is independence.
AI for your actual day: ten small wins most gonks miss
AI is most useful for small daily jobs like drafting replies, summarising docs, planning trips, and tidying spreadsheets, not just the big flashy corpo runs, choom.
Designing NWF: a feed format 62% smaller than JSON Feed
NWF is Neurowire's native feed format, roughly 62% smaller than the equivalent JSON Feed yet fully round-trippable, running interning, relative links, and date deltas. Lean chrome for any feed rig, choom.
Thirty seconds to fact-check any AI answer, choom
Jack into the cited source and check it actually says what the netrunner in the box claims, then cross-check one independent source, eyes hardest on numbers, dates, and names.
Why I chrome for the unremarkable grind, not the wow demo
A hundred engineers jacking into AI by default flatlines one jaw-dropping demo, choom. Production compounds on the thousandth unremarkable run, so measure throughput, not applause.
Neurowire Taps Pack: 271 curated sources, 24 themes
The Taps Pack is a ready-made rig of 271 vetted sources for Neurowire, sorted into 24 themes from Frontier AI Labs to Food. It's the curation layer running point like a fixer ahead of the incoming Neurowire SaaS, choom.
Neurowire: an open feed engine, now with full docs
Neurowire turns any source, even feed-less sites, into one canonical feed you can serialize to Atom, JSON Feed, Markdown, RSS, or its own compact NWF format, choom. The full docs rig is now live.
Why AI pilots flatline before they hit production
AI pilots flatline because they chrome up for a demo, not for adoption. Production needs architecture, ICE, and a crew that works different, not a fatter model.
What 700 chromed-up engineers taught me about adoption
After running 700+ crew through the training, the pattern's clear as neon, choom: adoption is a behaviour change, not a tooling drop. Role-specific reps and internal champions are what keep the rig running.