Ethan Westfall

The average person can comfortably speak 2 to 3+ times faster than they can type, yet most AI users (including vibe coders) still type rather than talk. That's why we've created VibeKeys Max, a tiny keyboard designed to make prompting effortless. vibekeys.dev/vibekeys-max.h…

I'm not sure how many people have stopped to consider the new economics of a token-usage business model, so here's a summary with tactical tips for reducing cost: We're in the midst of a fundamental shift in software economics: the transition from predictable per-seat SaaS licensing to variable, high-velocity token consumption. 🪙 The Economics of Tokens Traditional software costs are linear, but AI agent costs are exponential. When an engineer uses an agent like Claude Code, they aren't just paying for a "seat"; they are paying for the volume of data processed. - Context Window Compounding: Agents maintain history and retrieve large codebases to provide relevant answers. Each turn in a conversation sends more data than the last, causing costs to scale quadratically with the length of the task. - Recursive Workflows: High-end agents often spawn "sub-agents" to solve complex problems. A single human prompt can trigger dozens of automated API calls, each consuming thousands of tokens. - The "Success Trap": As shown by Uber’s 6x cost increase, the more helpful the AI is, the more engineers rely on it, leading to a budget "death spiral" where productivity gains are offset by massive inference bills. 🔓 Coping Mechanisms: The Shift to Hybrid Inference To prevent budget exhaustion, enterprises are moving away from a "Frontier-Model-Only" strategy toward Hybrid Inference. This approach optimizes for the "Price-Performance Frontier" by routing tasks based on complexity. 1. Local Inference (Edge Computing) Engineers are increasingly running smaller, open-weight models (like Llama 3 or Mistral) locally on their workstations. - Use Case: Simple autocomplete, syntax checking, and basic unit test generation. - Economic Impact: Zero marginal cost per token. It offloads the "low-value" high-volume noise from the expensive cloud providers. 2. Tiered Routing (Cheap Models) Instead of using a flagship model for every request, teams use "Router" logic to categorize tasks. - Small Models (e.g., Claude Haiku): Used for code reviews of simple diffs, summarizing documentation, or basic refactoring. - Large Models (e.g., Claude Opus): Reserved strictly for complex architectural decisions, cross-file debugging, or high-stakes logic. 3. Context Compression and Caching Companies are investing in "Prompt Caching," where frequently used parts of the codebase are stored in the provider's memory. This allows the model to "remember" the codebase without re-reading (and re-charging for) it every time a dev asks a question. In summary, for AI to be sustainable, the industry must move toward an orchestration layer that intelligently balances local compute for speed/cost and cloud compute for intelligence.

@Shivam25mishra Maybe there are other professions, but as a SWE the whiplash you described is very familiar. 😂

Unpopular opinion : prove me wrong but software engineer is the only profession where within 5 mins you can feel both an idiot and a genius.

@HussainIbarra From what you've observed, from where do such founders get their motivation to behave as you described? I sometimes check those boxes, but not as consistently as I would like.

Patterns I noticed in successful people in my life: - They have a psychopathic sense of urgency - They are "working" all the time. Even when it doesn't look like it - They are delusionally optimistic about their goals - They have a very small circle of friends - They are interested in ideas more than people - They write more than anyone else - They have the ability to make things happen even when it seems impossible

@petergyang This is, ironically maybe, my greatest gripe with AI-assisted development. I get so distracted when the agent is working that I lose focus on what I was working on / feel like I'm trapped in a malaise of sleepwalking at work.

How do people even do AI hackathons these days you're just sitting around waiting for the agents half of the time?

In that spirit, I invite those of you who enjoy vibe coding to check out a project I've been working on! An ease-to-use vibe-coding-specific secondary keyboard. Decent sales so far, so it must be meeting a real need! landing page: vibekeys.dev/vibekeys-max.h… user guide: vibekeys.dev/user-guide.html

Shamelessly promote your work. Shamelessly promote your ideas. Shamelessly promote your offers. Shamelessly promote your content. Shamelessly promote your products. Nobody else will do it for you.

@rxhit05 In the best products I use, it's actually something else: a solved (or at least greatly mitigated) problem. Everything you listed comes into play as well, but only after I'm convinced a problem I have is being addressed in a helpful manner.

What actually makes users stay? -product quality -habit formation -community -switching costs

@BacLeodiv camp in the mountains

Name one thing you still do better than AI.

@plainionist I built it on my corporate laptop, so I can't unfortunately. If I build one on my personal laptop at some point I can definitely share!

@EthanWestfall2 Interesting - do you mind sharing this skill with the community?

Programming with AI isn’t cheating. It’s a serious skill. 😎

@Aevmorfop Agreed. I personally do not like shuffling for very long 😆

@EthanWestfall2 True... But now there is just so much more of it. Is it worth shuffling through all the slop? Sometimes yes, sometimes no. Depends on how you feel on that day I guess 😅

At work today a really interesting presentation was shared about overcoming the shame associated with giving Claude credit for work well done. Ironically, the presentation felt very much like it was Claude generated. To complicate matters, the following presenter (a senior dev at Google) began his presentation by praising the previous talk for being "so real". He said this with sincerity, and I couldn't tell whether he realized he was praising a deck produced by AI. The thing I'm grappling with is whether my instinct that AI-generated content is not "real" in the same way as human-generated content is even correct. In some ways, maybe AI content co-signed by a human is more real because it helps the human discover what he believes more than he could himself. That doesn't seem quite right, so I'd appreciate others' insights.

Most people give the agent(s) one prompt and the first attempt is wrong (like you pointed out). The self correction capability is setting up a loop where the agent continues to test and fix its code until it's right. It can take many attempts, almost like trial and error, but eventually the agent gets to a solution that satisfies the PRD / specs.

@EthanWestfall2 @Umesh__digital If it can 'self correct' , why is it writing the wrong code in the first place..

Can someone explain to me why Anthropic's CEO keeps saying Software Engineering is dead, yet his company is still hiring Software Engineers?

@emironic @sflorimm this user guide has more info: #getting-started" target="_blank" rel="nofollow noopener">vibekeys.dev/user-guide.htm… the knob on the version you see in the user guide is for scrolling through options in conversations with an agent

@EthanWestfall2 @sflorimm what's the rotating knob is for?

Hey founders 👋 If your product solves a real problem, drop it here and tell us what problem it solves. I’ll rate the usefulness of as many as I can. Do the same. Founders support founders. Deal? 🤝

@S_N_SH_E_ Mostly just various claude models and gemini. Very rarely I'll dust of ChatGPT for good old times.

Be honest: How many AI tools do you still use daily after the initial hype wore off?

@Jbm_dev Hunting for product-market fit. Seems to be working to a degree. My team has tried so many different products (mostly failures), with a few actually doing well (e.g., vibekeys.dev/vibekeys-max.h…).

why are you building what your building? not the polished answer. the real one.

@karrisaarinen Agreed! There's a reason my favorite CS book is Clean Code.

What if we built better software, not just more of it?

@lswoogie Here is my AI tool (surprisingly it's hardware): vibekeys.dev/vibekeys-max.h…

Are you one of these people? - in tech - building your own AI product & learning AI tools - deeply into your own hobby - mine is photography - building a healthy daily routine Then you’re exactly the kind of person I’m looking to connect with. Let’s connect!

@thedevchandra most of the best engineers I know hardly ever get on social media -- can't be a coincidence😅

if you're in tech, do less code and more content this is the only way guys.

@adahstwt I've heard that SEO-wise they're all comparable, but I definitely think web traffic views only .com as reliably legit. Too bad it seems like all the good .com domains are taken or exorbitantly priced. 🙃

be honest... which domain instantly makes a startup feel more legit?👇 1) .ai 2) .com 3) .io 4) .app 5) .dev 6) .cloud 7) .sh