EngineerForAI

Just use the sdlc-bot. Make initiatives, throw them in the backlog. PO and scrum master agents manage the board, dedicated planning agents plan, generate tasks, pushed along the board, dedicated dev agents, pr manager agents, review agents get the work done. All running from peer connected AI Cli subscriptions.

I did it. I figured out the perfect AI coding loop It's called the Finn Loop and it's 100x'd my vibe code output while cutting the work I need to do down by 95% ANYONE can implement this. It's super easy. Here's how it works: 1. 3 new skills inside your coding agent. /spec, /build, /review 2. You start out with /spec "your idea". The skill then asks you enough questions to fully understand your idea in detail then create a detailed spec inside Linear (amazing project management tool. Very generous free tier) 3. In another Codex/Claude Code session: you have a /build loop running. Once a spec is created, the /build skill picks it up and builds it out for you. Then advances the status of the issue in Linear 4. In another session, you have /review loop running. It picks up every issue that is built out then reviews it. Checks for security issues, optimization issues, then tests the code in its own browser, lays out test steps, takes screenshots, creates a PR, and puts it in its own vercel test sandbox for you to test it 5. The review loop then takes all of that and pings you in a channel in your favorite messaging service. I use Slack for this. It shows you the PR, gives you all the testing steps, and executive summary of the changes, and a link to the Vercel sandbox where you can test it 6. You go to the message in Slack, review the change, then send a rocket ship emoji as a reaction. This signals to the loop that the PR should be merged. The only manual steps in this loop are you submitting an idea to your agent, and verifying the change at the end. Everything in between is automated in the loop. Your vibe coding workflow today is hand holding your agent step by step on everything you build. You spend all day doing this. Your vibe coding workflow when using the Finn Loop is you wake up, make a cup of coffee, give your agent a list of ideas, then come back at night and review all the changes. You're free to do whatever you want during the day. Let me know if you want my full spec, build, and review skills. Will release them soon. Will also film a tutorial video on this. There's also a ton more details I'll share shortly too, like the interface I built for monitoring all of this. You also can probably just send this tweet to your agent and it will recreate the Finn loop for you. Carve off some time and try this out today. I promise it will dramatically improve your workflow

What do you use AI for? In general I use it to produce revenue generating applications and do to do work I am paid for. Those applications and work are all hosted and delivered through cloud services. There are no "free models". There are self hosted models and if they work for you then great. They make no economic sense for me which is the point. I'm well aware that we live in a manufactured surveillance state. I've lived outside and inside that state for 30 years. Working with EFF and other orgs predating it. I get it. Inside the surveillance state, where I make money and benefit from it's fruit I use hosted and cloud services because it makes financial sense. If providers hobble their services to the point where it doesn't then I will stop using their services. Think for a moment though. If there are model setups that are cheaper and can compete on service then there is financial opportunity for someone to provide that to others for cheap. That's how it works. Naive people shout about buying surveillance state hardware like it's some kind of money saving prospect. It's not. And sure, if less compute and privacy are your priorities then you might satisfy some use cases. Show me the math if you want to bother. I've run the numbers. Btw all the "free" models are hosted as well and are far cheaper long run than paying upfront 10s of thousands of dollars that could otherwise be invested and earning you interest. Hardware is a depreciating asset and requires a lot of capital to make it make sense long term. I have personal computers of course. It's the cost of doing business. And you can rest assured that the power/value of ALL of my computers is inline with maximum capital retention. Paying for the latest at home compute has NEVER made financial sense. All the data is available.

'maximum intelligence' is a myth when you have zero agency. Providers can hike prices, gatekeep access, or silently lobotomize the model overnight (just think of how many times OpenAI and Anthropic have completely changed their models from one day to another). Enjoy renting a volatile black box that breaks your workflow every time they push a safety patch. Also, you will need to use some form of computer to use the "hosted compute" too, right? Are you going to pick that up from a landfill or what? Hardware buys control, plus you have infinitely more choice's, just go to huggingface.co/models and you'll realize how much more and better options you get access to for free.

You are going to be able to run Fable 5 locally on your desk In 2 years Apple will be coming out with Mac Studios with 1.5TB of memory With just 300gb of memory you can run Opus 4.8 level intelligence Think of what you can do with 5x that You need to be preparing for this now Start getting familiar with local AI technology Go to your Hermes/OpenClaw and use this prompt: “I am brand new to local AI and want to get familiar. Look at the computer you are currently on. Understand the specs. Then go on Huggingface and find the best models I can run on it. Then, walk me through how these models work, how they will run locally, and use cases I can do with them. After walking me through all of that so I’m educated, you can then load it onto this computer and build an interface so I can use them” In 2 years EVERYONE on Earth will have a local model running on their desk The people who start preparing now will be WAY ahead of everyone else

@2happyCSGO @AlexFinn @Dhaval @grok Let me simplify this. It will make more economic sense to pay for hosted compute vs buying hardware if the goal is maximum intelligence per dollar. In other words, more intelligence for less money. 30k Mac is not going to get you frontier intelligence for cheaper than hosted/api.

@engineerforai - is your handle serious or troll acc? You completely missed the nuance. 'Scaling via MoE' means the active parameters during inference stay low while the total model size grows. Ironically, this makes a high-RAM Apple Silicon Mac the absolute perfect machine for future models, because it has the massive unified memory needed to hold all those MoE weights without requiring insane compute for every single token. Grok literally confirmed my point: 'dense parameter gains show clear diminishing returns.' To use a hardware analogy you might actually grasp: if dense parameters = CPU clock speed (which hit a wall at about 6GHz), MoE is just adding more cores. We haven't moved towards CPU's with higher clock speeds since the 14900KS, we're just getting more cores. Active parameters per token stay low. Thanks for proving my point!

@2happyCSGO @AlexFinn @Dhaval @grok has the amount of parameters you can put in AI models plateaued and has the cost, speed and capabilites plateaued as well?

@engineerforai @AlexFinn @Dhaval the amount of params you can put into AI models have already plateaued completely. That's why 2026 is all about agents and not about "we now have this 50 trillion model ready"

Ai is saying that: Five advances likely to overcome the AI inference “memory wall”: Ultra-low-bit quantization — FP4/INT4 and below Shrinks model weights and KV caches, allowing more parameters to be transferred per second and held in memory. � NVIDIA +1 Speculative and multi-token decoding Predicts several tokens and verifies them together, turning otherwise idle compute into higher throughput while amortizing memory reads. � arXiv +1 KV-cache compression and improved attention architectures Quantization, token selection, latent attention and cache reuse reduce the rapidly growing memory required for long contexts. � arXiv +1 Faster, larger and more tightly integrated memory HBM3E/HBM4, 3D stacking, coherent CPU–GPU memory and faster interconnects increase capacity and bandwidth close to the processors. Current Blackwell Ultra GPUs reach up to 288 GB of HBM3E each. � NVIDIA Developer +1 Processing-in-memory and near-memory computing Moves matrix calculations closer to where weights are stored, greatly reducing the energy and time spent moving data between memory and compute units. � arXiv +1 Potentially most transformative: processing-in-memory. Most immediately impactful: quantization, KV-cache improvements and higher-bandwidth HBM.

@engineerforai @AlexFinn How you saying that? How much data centers you think thay have , I think it will be 200$/m

@AlexFinn @Dhaval You should consider the models that will be available in 2 years that won't run on a 30k Mac. And also consider the cost of a fable level model in 2 years. Buying your own hardware will never make financial sense.

@Dhaval Your brain still can't comprehend a world of abundance. That's fine. It will soon.

@surim0n Toronto

toronto has a lot of 🇨🇦 founders and no group chat. fixing that. slack/whatsapp group for people building. share progress, and most importantly UNBLOCK each other first happy hour once there's enough of us. what you also get: office hours from a claude & cognition ambassador reply "toronto"

@bryan_johnson If you rely on people's opinions, perspectives then you will succumb to them. Don't assume the identity of people that failed.

The world wants me to die. My incurable disease diagnosis became global news. It was omnipresent on social media and 1,900 articles were written in a matter of days. Many were saddened. However, joy dominated the commentary. People pointed to schadenfreude, the pleasure of another's failure. Yes, there’s that. There is a special place in people’s hearts that loves to see others fail, especially when that person’s presence threatens their own psychological stability in some way or helps them feel better about themselves. But, if you look over the social media commentary about me, you’ll see that pattern: “he deserved it.” I deserved it because I challenged death. The crowd was running a deeply rooted psychological script that represents the oldest, most deeply embedded stories of human culture. This was the first story ever written down, 4,000 years ago. Gilgamesh sought eternal life after losing someone he loved, only to have the plant of youth stolen by a serpent as he bathed. Leaving him to accept his mortality. Asclepius became so skilled at rejuvenation that he raised the dead. As punishment, Zeus struck him down with a thunderbolt to enforce life and death authority. This is the story of Jesus. Pontius Pilate offered a choice between a thief and the immortalist, and the crowd demanded the execution. People need this story conclusion to keep themselves sane. The challenger must lose and the loss must appear deserved. It’s a shield of self preservation. For if death is inevitable, their existence and that of their loved ones is justified and unavoidable. If death is not inevitable, nothing about their reality is safe. I occupy the same philosophical and archetypal position as Gilgamesh, Asclepius and Jesus. This statement will draw outrage and accusations of blasphemy, hubris and narcissism. Nevertheless, it’s the pattern that has repeated itself for thousands of years. Death has been the omnipresent concern of the human race. It encapsulates our greatest fears, joy and curiosities. The discourse around it changes over time; however, the fundamentals remain unchanged. What’s different about this moment, that is unlike any other moment, is that physical death may no longer be inevitable. What if I didn’t deserve it? And what if I am your ally, and not a threat?

@babayagatwt Google makes a lot of money. Anthropic and Open AI don't. Anthropic and Open AI's innovations will be swallowed by Google when it makes sense.

I still can’t figure out why Gemini struggles to compete with Claude and GPT. - Owns Chrome - Backed by Android - Stores most search results - Holds ~95% search history - Google has the biggest user data - Even incognito data isn’t fully private So what’s the problem?

@svpino Have you tried claude code? It's a great app. Claude desktop too.

I keep hearing that building with AI is easier than ever before, and yet I don't see that many new good applications? Why haven't I seen a significant increase in quality in the apps I use every day? Why aren't they bursting with new and useful features? Even the apps released by the labs that claim software engineering is dead are full of bugs and kind of meh. What gives?

@davj Try sonnet, fail fast.

Fable 5 is absolutely unreal. I gave it a startup idea, and in 8 hours it interviewed 100 users, built the v1, raised a 200 million seed round from a16z, blew the entire seed round on expensive GTM hires and slop ad campaigns that didn't convert, panicked and raised a down round, fired their entire leadership team, pivoted twice and then ultimately shut down the startup

@theo Did codex cli stop working for people or something? Works fine for me.

fwiw, I think that turning Codex into ChatGPT Desktop is a generational fumble

@weswinder Stop directly using anything and let your harness do it right in whatever way is best. Queues, loops, subagents...

stop using loops start using queues trust me

@ziwenxu_ No cli??

Codex is gone, now it's called ChatGPT Work.

@ziwenxu_ whatever framework that is.. opus recently used for me.

One prompt: 'build me San Francisco as a playable open world action game.' GPT-5.6 SOL Ultra did this.

This is what you want to see every week right?

@ziwenxu_ Do you know if we can use grok with opencode or pi? With x subscription?

just tested Grok 4.5 on cursor with the questions that break Opus (but rephrase it differently): - how many r’s in strawberry - is 5.11 bigger than 5.9 - the classic “car wash” / “cars vs wash” one - how many Aces left in the deck (new) It's safe to say Grok is very powerful

@kapilansh_twt Yes

as a vibe coder which plan are you actually on right now - Claude $20 - Claude $100 - Codex $20 - Codex $200 - something else be honest

Ohh I can't DM you. Anyway you need to download desktop app claude-sdlc-bot-web.vercel.app/settings/deskt… . Not sure if the macos release works. Linux/WSL does. I am pushing new release soon. Once you have desktop app walk through the wizard, configure your project. Start the poller. Open an claude/codex session and ask it to add an initiative to the backlog, some work you want done. Then move to Planning and tasks and dependencies, implementation plan is created. Once "Planning Done" move to "Ready" and then the bot will manage the implementation and PRs/Reviews etc. tickets will move across the board. If you add team members they connect to your project, run their poller and the bot on their desktop will pick up tasks, review PRs, plan initiatives using their AI cli subscriptions locally.

@engineerforai @16vchq Am I working with multiple agent engineer personas? Or is this for my existing human engineering team to collaborate? I tried using it, doesn't seem to be working.

Pitch what you're building. We invest $100K to $2M in ambitious founders through our Fellowship, Studio, Accelerator, and Fund.