![\"Pinboard](\"\/img\/pinboard-post\/board-in-progress.png\")
<\/p>\n\n
What\u2019s a mod?<\/h2>\n\n
Mods<\/a> are a new kind of Claude Code plugin: JavaScript or TypeScript that runs inside Claude Code and reacts to events like tool calls, prompts, or parts of the interface being drawn. Settings hooks, skills and MCP servers can\u2019t draw anything. A mod can, which is what made a sidebar possible. One caveat from the docs worth repeating: mods run with your permissions and aren\u2019t sandboxed, so only install ones you trust.<\/p>\n\n The board has three lists. Open decisions are questions waiting on me. Todos are the session\u2019s task list. Links are URLs from things Claude just made, like a PR, an issue or a push. I only capture links from commands that create something, because reads and test runs mention URLs constantly and I don\u2019t care about those.<\/p>\n\n Claude updates the board through a tool, and each update shows up as one dim line in the transcript, like The pane opens on its own the first time something lands on the board, if the terminal is wide enough. Otherwise Mods are new, and the API is still moving between releases, so I expect to be patching this.<\/p>\n\n What I keep noticing is how much of working with an agent is about attention, not capability. Claude could always keep track of my todos. I just couldn\u2019t see them.<\/p>\n","pubDate":"Sat, 03 Oct 2026 15:00:00 -0400","link":"https:\/\/jeradbitner.com\/blog\/pinboard-claude-code-mod","guid":"https:\/\/jeradbitner.com\/blog\/pinboard-claude-code-mod"},{"title":"A Defect or a Betrayal: What Jesus Actually Said About Sin","description":" At our Bible study last Monday, someone in the group challenged my whole way of thinking with one sentence: \u201cI\u2019m not sure I believe in sin.\u201d<\/p>\n\n\n\n I want to start by thanking him for that. Bringing up a core tenet and putting real pressure on it is harder than agreeing, and it got me thinking all week. I haven\u2019t been able to put the thought down. So here\u2019s where the study led me.<\/p>\n\n I was trimming my grapes the other evening, listening to K-Love on Pandora, when the thought arrived: if there is no sin, then what did Jesus die for?<\/p>\n\n That sent me looking at what the Bible actually says about sin, and what Jesus himself said about why he would die.<\/p>\n\n In 1 Corinthians 15:3, Paul passes along what he calls received tradition:<\/p>\n\n \u201cFor I delivered unto you first of all that which I also received, how that Christ died for our sins according to the scriptures.\u201d<\/p>\n<\/blockquote>\n\n So we learn this early in church and receive it as a core tenet. Fine. But received isn\u2019t the same as examined.<\/p>\n\n The other central passages line up the same way. Romans 5:8, God shows his love in that Christ died for us while we were still sinners. 2 Corinthians 5:21, the one who knew no sin was made sin for our sake. 1 Peter 2:24, he bore our sins in his body on the cross. 1 John 2:2, he is the atoning sacrifice for our sins and for the whole world\u2019s. Hebrews 9 and 10 run an extended argument that his death was a single, final sacrifice replacing the temple system. And Isaiah 53 is the Old Testament passage the early church read as prefiguring all of it, where the servant is wounded for others\u2019 transgressions.<\/p>\n\n Except those are all other people talking about sin. What did Jesus say?<\/p>\n\n From Jesus himself the material is sparser, but pointed. Mark 10:45, he came to give his life as a ransom for many. And the words over the cup at the Last Supper, Mark 14:24, blood poured out for many.<\/p>\n\n He says less about sin in the abstract than about specific people caught in it. But a few consistent moves run through the teaching.<\/p>\n\n He relocates it from the act to the interior. The Sermon on the Mount keeps taking a command and driving it inward: you\u2019ve heard do not murder, but anger already belongs to the same family; you\u2019ve heard do not commit adultery, but the look that possesses someone has already done the thing (Matthew 5:21\u201328). In Mark 7 he makes it explicit. Nothing entering a person defiles them. What comes out of the heart does. That one saying dismantles a whole purity system. Sin is something you generate, not something you catch.<\/p>\n\n He treats it as a condition producing symptoms rather than a tally of incidents. A tree is known by its fruit, and a good person brings good things out of the heart\u2019s storehouse (Matthew 7:17\u201320, 12:33\u201335). And in John 8:34, sharply: everyone who sins is a slave to sin. Not \u201chas done something wrong\u201d but \u201cis held by something.\u201d<\/p>\n\n He inverts the expected ranking, and this is the most disruptive part. The people in genuine danger, in his telling, are not the notorious ones. They\u2019re the ones certain they\u2019re fine. The tax collector who won\u2019t lift his eyes goes home justified; the Pharisee reciting his virtues doesn\u2019t (Luke 18:9\u201314). Those who are well have no need of a physician (Mark 2:17). Take the log out of your own eye first (Matthew 7:3\u20135). In John 9:41 he tells people that claiming to see is precisely what keeps their guilt in place. Admitted sin is workable. Unadmitted sin is the trap.<\/p>\n\n He isn\u2019t lenient about it, either. The same teacher who eats with the wrong people says it would be better to lose a hand or an eye than to be ruined whole (Matthew 5:29\u201330), and that anyone who trips up one of the little ones would be better off drowned (Mark 9:42). He takes it more seriously than his critics did, not less.<\/p>\n\n And in practice, he forgives before anyone has earned it. He tells the paralytic his sins are forgiven before he heals him. The woman accused of adultery is sent off uncondemned but not unaddressed: go, and don\u2019t do this again.<\/p>\n\n None of that actually defines the thing. It establishes why Christians believe Jesus died for our sins. It doesn\u2019t say what sin is<\/em>.<\/p>\n\n The word in the biblical languages is mostly an archery metaphor. Chet<\/em> in Hebrew, hamartia<\/em> in Greek, both carrying the sense of missing the mark. That\u2019s the base image. Aiming at something and falling short. Which already smuggles in a claim: there\u2019s a mark to miss, some intended shape for a human life.<\/p>\n\n Past that, the tradition stacks several distinct ideas under the same word. There\u2019s transgression, the legal sense, breaking a command. There\u2019s rupture, sin as a broken relationship between a person and God and between people, which is what the prodigal son story runs on and what we spent most of Monday circling. There\u2019s sin as a condition rather than a set of acts, the way Paul writes about it in Romans 7 as a power inhabiting people so that he does what he doesn\u2019t want to do, and in Romans 5 as something that entered through one man and spread to everyone, which is the seedbed of what later got called original sin. And there\u2019s Augustine\u2019s framing, later developed by Aquinas, of disordered love: not desiring bad things, but desiring good things wrongly, too much, or in the wrong order.<\/p>\n\n The archery image threads through all of them. The claim isn\u2019t primarily that you\u2019ve broken rules. It\u2019s that you\u2019re not yet what you could be.<\/p>\n\n The texts establish that Jesus taught about sin coherently and at length. If his authority is granted, the case is closed. If it isn\u2019t, then what you\u2019ve established is what a first-century teacher believed, which is a different thing and a much smaller one.<\/p>\n\n What\u2019s more interesting to me is that a good deal of what he describes doesn\u2019t need theology to be recognizable. The log in your own eye. The certainty that insulates you from seeing yourself. The gap between what you intend and what you actually do. The way harm keeps propagating when it isn\u2019t forgiven. Those are observable, and people with no religious commitment at all describe them constantly under other names. Someone can accept every bit of that and still resist the word \u201csin,\u201d because the word carries an extra claim: that the mark being missed was set by God, and that the failure is against a person rather than merely a standard.<\/p>\n\n The diagnosis lands whether or not you accept the physician.<\/p>\n\n Plainly, something is wrong with us. Not with circumstances, not with other people, but with us. Including me, and including you. Everyone who has looked honestly has found it. The thing I swore I\u2019d stop doing and did again. The ease of seeing another\u2019s fault against the difficulty of seeing my own. The way I can want to be good and still not manage it. Jesus described this more precisely than almost anyone, and he wasn\u2019t describing a doctrine. He was describing what he saw.<\/p>\n\n Whether you call it sin depends on one thing: whether there\u2019s someone the failure is against. Sin isn\u2019t just wrongdoing. It means the mark was set, and you owe an answer for missing it. Drop God and you still have the wrongdoing, but you lose the mark and the one who set it. You\u2019re left with a very good description of the human condition and no one to take it to.<\/p>\n\n That\u2019s the actual fork. Not whether the condition is real, because it plainly is. Whether it\u2019s a defect or a betrayal.<\/p>\n","pubDate":"Sat, 25 Jul 2026 00:00:00 -0400","link":"https:\/\/jeradbitner.com\/2026\/07\/a-defect-or-a-betrayal\/","guid":"https:\/\/jeradbitner.com\/2026\/07\/a-defect-or-a-betrayal\/"},{"title":"Showing my kids how phi and pi fit together","description":" At dinner last night, I got into a conversation with the kids about pi, phi, and sacred geometry. Which is not where every family dinner goes, but ours wandered there, and I was happy to follow it.<\/p>\n\n The hard part was that I could hear myself explaining something visual with too many words. Pi is easy enough to point at: circles, wheels, ripples in water, pizza if you need the obvious dinner-table example. Phi is stranger. It shows up as proportion, growth, spirals, pentagons, sunflowers, and that feeling certain shapes have where they seem balanced before you know why.<\/p>\n\n\n\n Then comes the question I was trying to answer: how do those two numbers relate to each other?<\/p>\n\n They feel like they belong in different drawers. Pi is the circle number. Phi is the golden ratio. One measures turning around a center. The other describes growth and proportion. But if you draw a regular pentagon and connect its corners into a five-pointed star, they suddenly show up in the same place.<\/p>\n\n That was the part I wanted the kids to be able to see.<\/p>\n\n So I made a small interactive page: Phi & Pi: Two Constants, One Geometry<\/a>. It walks through the golden ratio, circles, pentagrams, golden spirals, the Flower of Life, Metatron\u2019s Cube, and the Platonic solids. The goal is not to make a textbook. It\u2019s more like a little visual field trip.<\/p>\n\n Here\u2019s the relationship that still gets me:<\/p>\n\n That\u2019s not a poetic almost-connection. It\u2019s exact. The golden ratio can be written using pi because the regular pentagon divides a circle into five equal angles. Draw the diagonals, and those lines cut each other in the golden ratio. The circle gives you the angles. The star gives you the proportion.<\/p>\n\n Pi turns. Phi grows.<\/p>\n\n The golden spiral makes that easier to feel. Each quarter turn is a slice of a circle, so pi is doing the turning. Each step grows by the golden ratio, so phi is doing the scaling. The curve only works because both ideas are taking turns.<\/p>\n\n Sacred geometry sits in that same space for me. I don\u2019t think you need to treat every ancient diagram as a secret code for the universe. Some of that gets weird fast. But I do think there is something worth noticing when simple rules keep producing forms that humans across time have found beautiful, meaningful, and strangely familiar.<\/p>\n\n Overlapping circles make the Flower of Life. Connect the right points and you get Metatron\u2019s Cube. From there you can start seeing the shadows of the Platonic solids, and some of those solids bring phi right back into the room through pentagons and golden rectangles.<\/p>\n\n That\u2019s the thing I wanted to show the kids. Not \u201chere is a mystical proof,\u201d and not \u201chere is a dry math lesson,\u201d but: look at this. A circle, a five-pointed star, a spiral, a flower pattern. They are not random decorations. There is structure under them.<\/p>\n\n I like when math does that. It stops being a worksheet and becomes a way to notice the world.<\/p>\n\n And now, the next time dinner turns into a conversation about sacred geometry, I have something better than waving my hands around with a fork.<\/p>\n","pubDate":"Thu, 21 May 2026 00:00:00 -0400","link":"https:\/\/jeradbitner.com\/blog\/phi-and-pi","guid":"https:\/\/jeradbitner.com\/blog\/phi-and-pi"},{"title":"Forking My Claude Code Statusline","description":" About six months ago I wrote about dropping in hell0github\u2019s statusline<\/a> and calling it done. That stuck for a while. Then I kept wanting to change small things \u2014 the model abbreviation, which sections showed up, what the weekly bar actually counted \u2014 and So I forked it into its own repo<\/a> and changed how I install it.<\/p>\n\n\n\n The install script does one thing that matters: it symlinks If I\u2019m on a different machine I run the same two commands and I\u2019m done. The config lives next to the script, so it travels with the repo too.<\/p>\n\n The original was already doing the heavy lifting with The weekly percentage now reads from I added a 5-hour session block from Model names are abbreviated \u2014 There\u2019s an Every section is toggleable in A lot of glanceable signal in one line. The pipe between the cost bar\u2019s filled portion and empty portion is the projection \u2014 where ccusage thinks I\u2019ll end up by the end of the 5-hour block \u2014 colored by which layer it lands in.<\/p>\n\n Honestly, because the changes are opinionated. The hardcoded weekly limit was a real bug, but the model-name abbreviation, the effort section, the rearrangement of what\u2019s a \u201csection\u201d \u2014 that\u2019s all personal taste. It\u2019s easier to maintain my own version than to argue every preference into someone else\u2019s repo. And the symlink workflow means my taste is one The repo is at sirkitree\/claude-statusline<\/a> if you want to grab it or rip pieces out. Credit where it\u2019s due \u2014 the bones are still hell0github\u2019s, and the cost data still flows through ccusage<\/a>.<\/p>\n","pubDate":"Mon, 27 Apr 2026 10:00:00 -0400","link":"https:\/\/jeradbitner.com\/blog\/claude-statusline-fork","guid":"https:\/\/jeradbitner.com\/blog\/claude-statusline-fork"},{"title":"Meet the cast of Quantum Weave","description":" The Quantum Weave<\/a> site has had a magic-system page and a stack of short stories around it for months. What it didn\u2019t have was people.<\/p>\n\n That bothered me. The infographic<\/a> explained how the nine disciplines work, the narratives<\/a> showed reality fracturing in interesting ways, but if you landed on the page cold you couldn\u2019t tell who any of the stories were about, who they were trying to save, or who they were trying to stop. The world had nine sciences and zero faces.<\/p>\n\n Now there are eight. Mira Caldwell<\/a>, the master healer who hums old songs out of Veras while she works. Silas Greythorne<\/a>, the rogue architect with a publicly disastrous expulsion behind him and a charisma that still pulls people in. Yara Venn<\/a>, nineteen and self-taught, with light that obeys her even though her form is wrong. Vex Mortain<\/a>, the Whisper-rank operative whose mental signature changes color with the room. Thera Moss<\/a>, barefoot in the Highland Farmsteads with soil under her nails, slow and patient and immovable. Rowan Marrick<\/a>, late-awakened at twenty-six, asking questions of pre-collapse hinges that should not still hold. Torin Ashwright<\/a>, the ninth-generation forge master who sings when the work runs long. Kael Theron<\/a>, most-cited Weaver researcher alive, ink-stained, narrating his own hands.<\/p>\n\n\n\n I\u2019d been keeping detailed character files in a separate worldbuilding repo for a while. Three or four hundred lines apiece, full of magical profile, signature techniques, motivations, fears, story potential. Useful for me, useless for anyone else. The files were a stat sheet wearing a story\u2019s clothes.<\/p>\n\n So I tried something. What if the public page read like introducing the character to a close friend? Less D&D-block, more \u201clet me tell you about this person.\u201d Not \u201cMira Caldwell, age 42, she\/her, primary disciplines Corpus and Luxomancy at master level,\u201d but instead, \u201cIf you walked into Mira\u2019s ward at the Restoration Academy on any given afternoon, you\u2019d find her humming.\u201d<\/p>\n\n That second framing wants different things from the prose. It wants a present-tense scene to open. It wants specifics over abstractions \u2014 a Veras Market collapse, a particular braid, the way her phrasing falls in pitch when she\u2019s calming a patient who doesn\u2019t realize they\u2019re being calmed. It wants the writer to sound like they actually know the person.<\/p>\n\n The tradeoff was harder than I expected. The first draft had spoilers all over it because the most compelling material I had was exactly the stuff a reader shouldn\u2019t see before chapter one \u2014 active investigations, hidden truths the character doesn\u2019t yet know, the names of antagonists who matter to a later arc, the specific traumas that drive each story. Each profile got revised to focus on who the person is<\/em> publicly, what they do<\/em> for a living, what their reputation says about them. Less story, more dossier-with-a-pulse.<\/p>\n\n Each profile has a full-size portrait on the left that stays visible as you scroll, with the narrative on the right. Discipline names in the prose get a small dotted underline; hover one and you get a tooltip with the discipline code, a one-line summary, and a link to the full discipline page. Other characters mentioned in a profile become links to their own profile, with a tiny hover card showing name and role.<\/p>\n\n That cross-linking was a discovery. Once I had all eight profiles in front of me I noticed the characters mention each other constantly. Mira and Kael published together. Kael\u2019s peer review touched Silas\u2019s expulsion. Vex has worked the Forge Masters. Rowan trades custom tools with Torin. The character pages map onto each other the way disciplines do, and letting a reader click through that web feels right for a worldbuilding site.<\/p>\n\n The whole pass was a single commit<\/a>, with a generous assist from parallel Claude Code agents \u2014 eight in parallel for the first draft of the profiles, then eight more for the spoiler-strip pass. I sat in the conductor seat, the agents handled the typing.<\/p>\n\n I don\u2019t know yet how the dossier voice will hold up across 100,000 words of finished novel. The profiles work as standalone introductions, the way an appendix sits at the back of a book, but the voice I used to write about<\/em> the characters isn\u2019t the voice I\u2019ll use to write with<\/em> them. Mira\u2019s dry humor on her profile page is a description. In a scene she\u2019ll have to do it.<\/p>\n\n There\u2019s also the question of whether the portraits I\u2019m using are the ones I want to print. They were generated, refined, regenerated, and at some point I called them done because I had to call them something done. If I ever actually publish I\u2019ll probably commission proper art. For now the wiki has faces.<\/p>\n\n The character index<\/a> is the door if you want to poke around. Mira and Torin and Rowan are the main cast<\/a> at the top; the other five orbit them. The infographic<\/a> explains what the disciplines do, and the narratives tag<\/a> has the short stories that started this whole thing.<\/p>\n","pubDate":"Sun, 26 Apr 2026 00:00:00 -0400","link":"https:\/\/jeradbitner.com\/blog\/meet-the-cast-of-quantum-weave","guid":"https:\/\/jeradbitner.com\/blog\/meet-the-cast-of-quantum-weave"},{"title":"Jenney's Story Featured on the Cancer Research Institute","description":" The Cancer Research Institute published Jenney\u2019s patient story<\/a> on their site. It\u2019s strange seeing our worst year turned into someone else\u2019s summary.<\/p>\n\n\n\n If you\u2019ve been reading here a while, you already know the broad strokes \u2014 headaches during pregnancy, two brain surgeries, stage 4 metastatic melanoma while we had a newborn and four other kids at home. The CRI piece covers the medical side well: the immunotherapy treatments, tumors disappearing after one infusion, the side effects that forced treatment to stop early, and the October 2020 scan that came back clean.<\/p>\n\n What a patient profile can\u2019t really get at is the rest of it. The 3 AM conversations. The way our older kids stepped up in ways no child should have to. Watching Jenney relearn our children\u2019s names after brain surgery \u2014 not as a metaphor, but literally, one syllable at a time.<\/p>\n\n That\u2019s what I wrote the memoir about. Learning Your Children\u2019s Names Again<\/a> has the parts that don\u2019t fit in a medical profile \u2014 the fear, the dark humor, the weird mundane details that somehow matter more than the scans did.<\/p>\n\n If you read Jenney\u2019s story on CRI and want the full picture, the book is available now:\nKindle eBook<\/a>\nPaperback<\/a><\/p>\n\n Jenney\u2019s still in remission. The kids are getting big. Some weeks I forget any of it happened, which I think is its own kind of miracle.<\/p>\n","pubDate":"Thu, 16 Apr 2026 00:00:00 -0400","link":"https:\/\/jeradbitner.com\/2026\/04\/jenneys-story-on-cancer-research-institute\/","guid":"https:\/\/jeradbitner.com\/2026\/04\/jenneys-story-on-cancer-research-institute\/"},{"title":"Designing with Pencil and Claude Code","description":" I\u2019ve been building two internal tools at Lullabot lately. OpptyNet finds government RFPs for us. Upstraker tracks upstream patches across our open source contributions. Both needed real UI design, not just \u201cmake it work\u201d design. And I found myself doing that design work almost entirely through conversation.<\/p>\n\n\n\n The tool is called Pencil<\/a>. It\u2019s a design editor that integrates with Claude Code through MCP, so you can describe what you want, iterate on layouts, build component libraries, and screenshot your work without leaving the terminal. The I was skeptical. But after running two projects through it, I\u2019m genuinely impressed with how well it works.<\/p>\n\n With OpptyNet, I made the classic mistake. I jumped straight to pixels. Back in late March I had Claude whip up some UI brainstorms, an opportunity detail page, a dashboard list view, and they looked fine. Plausible even. But they were decorating a house before pouring the foundation.<\/p>\n\n So I backed up. I pulled out the Ponchot methodology, a lean design process we use at Lullabot, and actually did the work. Purpose definition first. Who is this for? What does success look like? The answer was specific: this is for Jenna, our business development person, and she needs to go from email digest to go\/no-go decision in under 60 seconds per opportunity. That\u2019s the whole job.<\/p>\n\n Then content modeling. Priority-driven requirements. A design system built on our existing Lullabot brand library. Then<\/em> pixels.<\/p>\n\n One of the more satisfying pieces of this project was building a proper Pencil design library for Lullabot. Fifty-six color variables. Typography tokens using Circular and IBM Plex Mono. A spacing scale on a 4px grid. Sixteen reusable components: buttons, cards, form inputs, nav links, badges, tags, blockquotes. The whole brand system in a OpptyNet imports it. Upstraker imports it. When I start the next internal tool, it\u2019ll import it too. Two projects in and the library is already saving me real time.<\/p>\n\n Once I had the foundation, the purpose statement, the content model, the design tokens, I started iterating in Pencil. Because the feedback loop is just conversation<\/em>, the iterations came fast.<\/p>\n\n V1 was card-based. A list of opportunity cards with a filter bar and pagination. Standard stuff, but too much visual weight per item. Jenna needs to scan 20-50 opportunities. Cards are too heavy for that.<\/p>\n\n V2 went dense, a table layout with inline score bars. Better for scanning. But it lost the warmth. Felt like a spreadsheet, and the whole point was to be better<\/em> than the spreadsheet she already has.<\/p>\n\n V3 tried a split panel, like an email client. List on the left, preview on the right. This was the layout that stuck. You scan on the left, evaluate on the right. No page transitions, no context switching.<\/p>\n\n V4 and V5 were mashups, combining the split panel with a dark hero banner for the selected opportunity, stat cards, score breakdowns alongside contacts. Getting closer, but cluttered.<\/p>\n\n Then I ran the Ponchot critique against V5. The stats were eating too much vertical space. There were no filters, which is a bad miss when your design principle is \u201cspreadsheet speed.\u201d The Ask AI feature, which is arguably the differentiator, was hidden in a tab.<\/p>\n\n V6 fixed all of that. Stats compressed to a single ribbon. Filter pills below the list header. Ask AI pinned to the bottom, always visible. Keywords gone because they were redundant with the score breakdown.<\/p>\n\n V7 was polish. Starbridge, the $15K\/year tool we\u2019re replacing, has this restrained luxury feel. So I added ambient glows on score badges. A subtle gradient on the dark left panel instead of flat color. Red glow on action buttons. Cards with soft shadows instead of borders. Nothing that changes the information architecture, just surfaces that feel crafted<\/em>.<\/p>\n\n Seven versions. One conversation thread. Maybe three hours of actual work. That last part still surprises me.<\/p>\n\n With Upstraker, I already had the Lullabot design library in place. The concepts came together faster because I wasn\u2019t building the visual language from scratch. Same brand tokens, same component patterns, different product.<\/p>\n\n Having a shared design library is weirdly freeing. You stop thinking about what shade of blue to use and start thinking about what the interface actually needs to do<\/em>.<\/p>\n\n The Ponchot process caught gaps that seven iterations of pixel-pushing wouldn\u2019t have found. The missing filters. The buried AI feature. Those were purpose problems, not visual problems. No amount of glow effects would\u2019ve fixed them. I keep learning this lesson and then forgetting it: methodology before pixels.<\/p>\n\n The speed of conversation-driven design caught me off guard too. When the iteration cycle is just \u201cthat\u2019s not quite right, try compressing the stats to one line,\u201d you cover ground fast. No switching between Figma and a chat window and a browser. The design and the code live in the same environment, which means they share context from the start.<\/p>\n\n I\u2019m not ready to say Pencil replaces a dedicated design tool in every case. But for internal tools and rapid prototyping, it handled everything I threw at it. And the design library I built for the first project made the second one noticeably easier. I\u2019m curious what the third one will feel like.<\/p>\n","pubDate":"Fri, 03 Apr 2026 00:00:00 -0400","link":"https:\/\/jeradbitner.com\/2026\/04\/designing-with-pencil-and-claude-code\/","guid":"https:\/\/jeradbitner.com\/2026\/04\/designing-with-pencil-and-claude-code\/"},{"title":"New Short Story: Tale of a Teenage Robot","description":" I published a new short story \u2013 Tale of a Teenage Robot<\/a>.<\/p>\n\n\n\n It started as a dream and turned into something I couldn\u2019t stop thinking about. A suburban dad finds a gecko wearing self-made robotic armor in the garden. It grows. It absorbs Legos, a wifi router, drill bits. It evolves through what feels a lot like childhood \u2013 clumsy and eager at first, then sullen, then hostile. Somewhere along the way it stops chirping and starts talking back.<\/p>\n\n I wrote it in the voice of a guy who manages projects for a living and tries to manage this the same way. It doesn\u2019t go well.<\/p>\n\nWhat it does<\/h2>\n\n
Pinboard: +4 todo, +1 decision<\/code>. The current board also gets appended to the system prompt on every request, so Claude always knows what\u2019s open without me reminding it. That also means Pinboard doesn\u2019t lean on TodoWrite<\/code> or the Task tools, which newer models don\u2019t get by default anyway. Claude Code\u2019s own task list isn\u2019t something I can count on being there, so the board brings its own.<\/p>\n\n\/pinboard<\/code> opens it. Everything resets on \/clear<\/code>, which is how I want it. It\u2019s a whiteboard for this conversation, not a project tracker.<\/p>\n\nTry it<\/h2>\n\n
claude plugin marketplace add sirkitree\/pinboard\nclaude plugin install <\/span>pinboard@pinboard\n<\/code><\/pre><\/div><\/div>\n\n\n
But what is it?<\/h2>\n\n
Where that leaves the original challenge<\/h2>\n\n
phi = 2 \u00d7 cos(pi \/ 5)\n<\/code><\/pre><\/div><\/div>\n\ncurl<\/code>-ing a fresh script into ~\/.claude\/<\/code> every time I tweaked something got old fast.<\/p>\n\nThe symlink trick<\/h2>\n\n
statusline.sh<\/code> and statusline-config.json<\/code> from the repo into ~\/.claude\/<\/code>, backing up whatever was there before. That means the file Claude Code runs is the file in my git checkout. Pull the repo, the statusline updates. No copy step. No \u201cwait, did I edit the right one?\u201d moment that I\u2019ve definitely had before with dotfiles.<\/p>\n\ngit clone git@github.com:sirkitree\/claude-statusline.git ~\/repos\/claude-statusline\ncd<\/span> ~\/repos\/claude-statusline &&<\/span> .\/install.sh\n<\/code><\/pre><\/div><\/div>\n\nWhat I actually changed in the script<\/h2>\n\n
ccusage<\/code> and the multi-layer cost bar. The changes that mattered to me:<\/p>\n\nrate_limits.seven_day.used_percentage<\/code> in the JSON Claude Code pipes to the statusline. The previous approach divided ccusage\u2019s dollar cost by a hardcoded $500 weekly cap for max5x, which is just wrong \u2014 Claude Max limits are message-based quotas, not dollar quotas. So my old statusline was confidently telling me \u201cweekly 0%\u201d when claude.ai\/settings\/usage said something else entirely. Now they agree.<\/p>\n\nrate_limits.five_hour<\/code> that shows the percentage and time until reset, which is a more useful signal than the ccusage block timer when you\u2019re trying to figure out whether to keep going or wait it out.<\/p>\n\nClaude Opus 4.7<\/code> becomes o4.7<\/code>, sonnet becomes s<\/code>, haiku h<\/code>. If the context window is 1M it tacks on (1m)<\/code>. Saves horizontal space, and at this point I recognize the shorthand faster than the full name.<\/p>\n\neffort<\/code> section now too, since some models expose \/effort<\/code> levels (low\/medium\/high\/xhigh\/max \u2192 lo<\/code>\/md<\/code>\/hi<\/code>\/xh<\/code>\/mx<\/code>). Useful for remembering what I set the dial to three hours ago.<\/p>\n\nstatusline-config.json<\/code> under sections.*<\/code>. Colors live there too. The thresholds for the multi-layer cost bar (where it flips green\u2192orange\u2192red) are configurable instead of hardcoded.<\/p>\n\nWhat it looks like now<\/h2>\n\n
sirkitree.github.com | o4.7 | mx | master | 23k\/200k [\u2588\u2588\u2591\u2591\u2591\u2591\u2591\u2591\u2591\u2591] | $12\/$140 [\u2588\u2502\u2591\u2591\u2591\u2591\u2591\u2591\u2591\u2591] 9% | weekly 34% | session 18% \u2192 2h 14m | \u00d71\n<\/code><\/pre><\/div><\/div>\n\nWhy a fork instead of a PR<\/h2>\n\n
git pull<\/code> away on every machine I use.<\/p>\n\nWriting them<\/h2>\n\n
The wiring<\/h2>\n\n
What I\u2019m still figuring out<\/h2>\n\n
.pen<\/code> files it produces are encrypted, only accessible through the Pencil tools themselves. You don\u2019t open them in Figma. You don\u2019t drag boxes around with a mouse. You talk about what you need and the design materializes.<\/p>\n\nStarting wrong (and backing up)<\/h2>\n\n
The Lullabot design library<\/h2>\n\n
.pen<\/code> file that any project can import.<\/p>\n\n![\"Lullabot](\"\/img\/pencil-post\/lullabot-design-library.png\")
\nThe Lullabot design library: color palettes, typography, spacing, and reusable components.<\/em><\/p>\n\nSeven versions of OpptyNet<\/h2>\n\n
![\"OpptyNet](\"\/img\/pencil-post\/opptynet-v1-cards.png\")
\nV1: opportunity cards with score badges, platform tags, and deadline countdowns. Looks nice, scans slow.<\/em><\/p>\n\n![\"OpptyNet](\"\/img\/pencil-post\/opptynet-v2-dense-table.png\")
\nV2: compact table rows with stat ribbon. Faster to scan, but sterile.<\/em><\/p>\n\n![\"OpptyNet](\"\/img\/pencil-post\/opptynet-v3-split-panel.png\")
\nV3: the split panel layout that stuck. List left, detail right.<\/em><\/p>\n\n![\"OpptyNet](\"\/img\/pencil-post\/opptynet-v4-mashup.png\")
\nV4: dark hero banner meets split panel. The layout is finding its shape.<\/em><\/p>\n\n![\"OpptyNet](\"\/img\/pencil-post\/opptynet-v5-final-mashup.png\")
\nV5: stats, contacts, notes, and AI chat all present. Too much competing for space.<\/em><\/p>\n\n![\"OpptyNet](\"\/img\/pencil-post\/opptynet-v6-refined.png\")
\nV6: filters added, stats compressed, Ask AI always visible at the bottom.<\/em><\/p>\n\n![\"OpptyNet](\"\/img\/pencil-post\/opptynet-v7-polished.png\")
\nV7: the final design. Ambient glows, gradient panel, soft shadows. Same bones as V6, better skin.<\/em><\/p>\n\nUpstraker: starting from the library<\/h2>\n\n
![\"Upstraker](\"\/img\/pencil-post\/upstraker-dashboard-light.png\")
\nUpstraker\u2019s contribution dashboard: issues close to landing, top projects by patches, and an org bump chart.<\/em><\/p>\n\n![\"Upstraker](\"\/img\/pencil-post\/upstraker-dashboard-dark.png\")
\nThe same dashboard in dark mode. Both themes came out of the same design library tokens.<\/em><\/p>\n\n![\"Upstraker](\"\/img\/pencil-post\/upstraker-dependency-issues-light.png\")
\nThe dependency issues view, tracking patches across all projects.<\/em><\/p>\n\nWhat I walked away with<\/h2>\n\n
![\"Raphael](\"\/assets\/books\/tobit\/raphael_revealed.jpg\")
<\/p>\n\n