First Serious Product Work with OpenCode Beats
This might be the inflection point where I attempt to do serious product work™ using OpenCode and a local model for the first time. Didn't have the bandwidth to experiment as Claude and Codex already exist. But this is alluring. It did the pelican better. And that IS a feat… https://t.co/XEE1ejhTMo
Quantizing LLMs with UnslothAI Supercharges Developer Productivity
Gotta say, @UnslothAI quantizing every interesting LLM into usable sizes has to be one of the highest-impact activities on global dev productivity this century. The fact that we can just grab models like https://t.co/PpDF1OYqk1 and _keep_ them is magical. Thanks, Dan &...
Developers Panic when Claude API Goes Offline
A million agents scream in pain, realizing their own futility whenever Claude's API is down. Can't even have Claude call Codex to do a code review right now. 🤣 How can people POSSIBLY get any work done like this? https://t.co/75n3jkNWYN
Price SaaS for Unpredictable Agent Counts, Not Limits
Take note, SaaS founders. People will run somewhere between 0 and a few dozen agents on your platforms. Per team. Per user. Per project. You won’t know. THEY won’t know. Pricing should facilitate this, not hinder it.
SaaS Is Becoming a Family‑run Legacy Business
I just realized that we are probably in the decade when the first meaningful number of long-term SaaS founders will turn their profitable businesses over to their children. It's pretty weird to think of SaaS as a family business. I...
Podscan Turns Podcast Mentions Into Instant Book Recommendations
Oooh, big use case for Podscan, and one that intersects my two favorite pastimes: reading and podcasts. Check out Simon's Podshelf, this is a genius approach to trans-media-referencing. Extracting book recommendation as soon as they're mentioned on shows.
LLMs Poised to Match Printing
In a couple decades, people will start to unironically compare LLMs to the invention of the printing press, the radio, and the internet. It is hard to underestimate the impact these thinking machines already have and will expand onto people’s baseline...
AI Platforms Will Monetize Aggressively, Driving Away Unwanted Users
Like it or not, this is the kind of fun monetization optimization strategy we can expect from AI platforms in the years to come. They'll turn every knob and fondle every lever passionately to ensure that unwanted users migrate off their...
Founders Choose: Cheap VPS Tinkering vs Managed Scaling
So, this is topping Hacker News, where an indie dev recommends running SQLite on lean VPSes forever. At the same time, @JackEllis speaks about using managed servers for critical production systems from day one. They're both true. Just at different points in...
Leadership, Not Hands‑on Work, Defines Founders and Engineers
Of course you’re a founder if your product is assembled by other people and agentic tools. You can even call yourself an engineer if you held the reins.
Claude Code's 2‑hour Promise Yields Barely a Prototype
People assume they can build any MVP with Claude Code in 2h. But what they build is a MP at best. Maybe even just an M.
AI Will Auto‑Label Pages for Skeptics' Comfort
Realistically, the result of this will be a “Hey Claude Code, add a Human Crafted label to all pages that feel like AI skeptics might land on” prompt. Not sure if this is funny, sad, inevitable, cunning, or else.
AI Orchestration Prioritizes Smart Token Use Over Cheap Speed
The delegate starts delegating. This is what I love about AI orchestration, and I thinks it’s only reasonable that we’re now entering the phase where token usage and model intelligence choices matter. I prefer this honest optimization to the subsidized...
AI Now Drafts Policies Reliably with Peer Review
And instead of having to hand-craft these policies, we now have magical thinking machines that can do it for us, reliably, and with peer review.
Git Worktrees Demand Automated Per‑branch DB and Redis Provisioning
Git worktrees are all the rage for even slightly parallelized agentic engineering. I wonder, how do you deal with "basic" external dev services? Who spins up a new db (w/ fresh data), or a Redis instance, for each worktree? Is this a...
Veteran Developers Boost Speed Using Agentic AI
If you need any more convincing that seasoned and highly capable developers are using agentic AI as a productivity tool without creating slop, here’s another example.
Do Whatever for $200: Developers' True Shopping Cart Test
I support this, both the angle and the change. "Do whatever for $200" is the ultimate shopping cart test for developers. If you've been burning tokens on performative agentic nonsense, you've actively contributed to this change.
Founders Pull Back Transparency as Complexity Rises
Got a lot of feedback on this one. Plenty of founders who once did (or planned to) build in public have significantly curtailed their transparency. When the idea/execution-complexity balance changes, marketing and expertise-signaling methods quickly follow.
AI Turns Public Transparency Into Competitor Blueprint
Building in public has been fundamentally changed by the advent of AI. Pains me to say it, but the transparency playbook that built my career now hands competitors a blueprint. Here's what I still share (and what I never will. Anymore.) https://t.co/dejGIMcP7M
Anthropic Leaked Claude Code for Covert Community Review
Hot take: Anthropic leaked Claude Code intentionally to get a nerdosphere code review it would have never gotten if they had just open-sourced it.
Payment Fees Are a Business Cost, Not a Complaint
The people surprised by this have yet to learn of the cost of doing business and just how much a payment provider does for them. $17k on $600k revenue isn’t something you complain about out. You price it in.
Avoiding Parallelism: One Task at a Time Improves Focus
I'm capped at 1. The moment I start juggling, I drop my concentration. An agent is a DIRECT execution tool for me, not a delegate-among-many. Might be a me issue, or a tooling issue. But right now, I'm staying away from...
Unearthing Hidden Claude Skills Through Random Mentions
The discovery of Claude skills still is an unsolved problem. I learn of the most insane high-impact skills from throwaway mentions in some reply somewhere. My mobile browser is full of links like these. What’s the most impressive Claude skill...
Expensive Tiers Boast, Real Builders Stay Silent
Yeah, this tracks. A lot of the people I see foaming at their mouths have been using the $200 tier for performative "here are my 10 parallel agents" screenshots and have very little to show for it. Meanwhile, the builders...
Slow, Low‑heat Onions with Butter Transform Every Dish
Biggest culinary upgrade for me in recent years was lowering the cooking temperature for onions and taking lots of time (and, of course, butter). Makes any and all dishes better. Can't rush aromatics.
Podscan Finally Mentioned Alongside Apple After Two Years
Obviously, I have set up alerting for "Podscan" on Podscan itself. And I love how podcasters are starting to mention the platform along with the massive incumbents like Apple. This one just came in. Took 2 years to get to this...
Proactive Code Scanning: Fix Bugs Before They Occur
Now extrapolate this into the future: instead of just reactively fixing issues when they occur, the next step is to proactively fix them as they are detected by a background agent that scans the codebase 24/7 and simulates error scenarios.
Need Git Commits to Capture Full Prompt Context
I wish Claude Code would automatically include the prompts + the full context in a repo as a git commit "note". The full conversation that led from zero to feature, and I want it in git blame, not just who did...
Future 10x Engineers Thrive on Code Review, Not Writing
The 10x engineer of tomorrow will look quite different from what it has been. I was never an outstanding dev. But I have a sense of what performant and good code looks like. Been copying it from SO for years.🤣 And now,...
AI Speeds up PHP Tests From 20 Minutes to 3
My full PHP test suite took 20+ minutes. Unbearable. Yesterday, I asked Claude Code to speed up my test suite. I knew that parallel testing would be a massive time saver, but I had set it up so that things immediately...
Claude's OS Control Makes OpenClaw Mac Minis Indispensable
All those OpenClaw Mac Minis are going to be really useful now that Claude has full control of the OS.
Edge AI Brings Private LLMs to Personal Devices
AI is moving to the edge. Private, isolated LLMs are coming to our tablets, phones, and laptops. Ownership, privacy, access. All things APIs don't/won't/can't give us now. This is significant. And sure, it's slow right now, but this is the slowest it...
AI Can Instantly Double Your Test Coverage
Okay this might be totally obvious but you can just tell Claude Code to double your test coverage and it will actually do it. Will need a lot of review, but if you’re just getting started with adding tests to your...
Ban on AI Tools Feels Like Losing Your Driver’s License
This kind of ban (on Claude, your Google account, or similar) is becoming the digital equivalent of losing your drivers license. And for what? Using some open source agentic harness because you saw someone share a cool screenshot? :/
90% of Queries Resolved by Knowledge Base Articles
Opposite experience, from the other side (operating one). The majority of user questions can be answered through surfacing the right knowledge base article or quoting from it. Remaining 10% come through. Nathan likely is a more “resourceful” customer who does actually...
AI Providers Clamp Down on Agents; Founders Unprepared
Google is banning accounts. Anthropic is locking down their plans. Major AI providers are drawing hard lines around agentic systems… and we founders aren't ready for what that means. Not just for our agentic coding, but how we deal with liability...
Never Manually Edit Code When Using AI Agents
One of my most painful realizations using agentic systems is that they treat all work-in-progress as ephemeral. They’ll gladly reset your branch or dev database to “try something”, because they expect to write working code eventually. Never manually add code while...
Transparent Ads Preserve Trust; Hidden Sponsorship Harms Credibility
“Native” advertisement, am I right. Ads are fine, as long as it’s clear they’re paid placements. Stuff like this breaks immersion and destroys author trust. Horrible practice.
Cut AI API Costs by Using Flex Tier First
If you do AI inference via OpenAI’s API, you should use the flex tier for half price. My requests always try to use flex tier first, and on 429 / 500 errors, I use the default service tier. 95% of my...
AI Overpromises: Claude Can't Deliver All in 30 Minutes
Claude, you'll be done with implementing ALL of these improvements in like 30 minutes, including full test coverage and security analysis. You know it. I know it. Which, as Claude shows pre-AI estimates, is WILD. https://t.co/aB2o472961
AI Makes Public Building Risky; Founders Turn Silent
This is why building in public, as fun and empowering as it is, has severe limits. Competitors have a significantly easier time now than before AI to build functionally critical-mass-escaping clones to swipe the bottom of the barrel. Lots of founders are...
AI Turns Vague Ideas Into Working Code Instantly
People acting like "Hey Claude Code, implement this feature I dreamt of just now and have no idea how to even start building" and then prototyping a fully-fledged working implementation isn't anything short of magic.
Achieving UI/API Parity Enables Rapid Custom Interfaces for All
This is a long-term bet of mine, which is why I’m working on full parity between UI functionality and API affordances. Because ultimately, this opens the door to full custom UIs spun up in a few agentic coding hours on top...
Prompt Claude Code for Full OWASP Security Sweep
What do you use to have Claude Code do a full security sweep of your codebase? My go-to is “Run a deep OWASP security sweep of the full app, all APIs and any internal services. Report in descending severity and suggest...
AI on CPU Will Become as Routine as JSON.parse
AI on CPU means AI-as-a-primitive. AI.answerQuestion() will be as normal as JSON.parse() once this becomes baked into either the frameworks, languages, or even on the OS level. And I’m very much here for it. Love to see these steps towards...
Walmart Urged to Secure Email Relay Against Phishing
Hey @walmart (or maybe @Walmarttech), could you kindly configure your email relay domain(!) so it can't send phishing emails? I get these once a week, and it's a pretty bad look. https://t.co/lerflUl2eR
Chrome Extension Becomes Malicious After Ownership Transfer
Roses are red. Violets are blue. Chrome Extension Turns Malicious After Ownership Transfer, Enabling Code Injection and Data Theft
Give SaaS Users an API, Empower Their Agents
Every SaaS must offer an API to its data & functionality. If you run a SaaS, consider this: eventually, all your serious users will run some kind of agentic tool to get work done. The question is: will you make it easier...
Never Let Claude Manage Production Infra without Humans
Unless you're Pieter Levels, do NOT RUN CLAUDE on your production system or infrastructure. It's a very charismatic toddler when it comes to understanding business consequences. It WILL do something completely disastrous eventually. Human in the loop for all infra work. ALWAYS.
Unresolved Technical Debt Undermines Product, Morale, and Business
Just like regular 0.8x developers, but this time much faster, unresolved technical debt will kill first any real improvements to the product, then morale, then the business.