BobCA
Your AI team that learns, thinks, and builds with you
46 followers
Your AI team that learns, thinks, and builds with you
46 followers
AI coding tools can write snippets, but they don’t work like a real team, you still prompt, fix, re-explain, and build alone. Bob’s Workshop introduces a new way to build: a cognitive AI team that learns, thinks, and builds with you. These aren’t one-off responses, but AI teammates with memory, reasoning, and context who collaborate across your workspace to plan, code, review, and ship together.
This is the 2nd launch from BobCA. View more
BobCA
Launching today
Most AI assistants are just glorified autocomplete. BobCA gives you a sovereign digital workforce. Using your learned engineering DNA, set grading targets in Mission Control, and let your digital twin autonomously debug, refactor, and build in the background. It logs negative constraints so it never repeats a mistake. Track every bug crushed and feature shipped live on the Autonomy HUD. Stop explaining your codebase and start managing your twin.
Free Options
Launch Team / Built With
BobCA
Hey Everyone👋 Kemone here, founder of Bob's Workshop.
What is BobCA? BobCA (Bob Code Assistant) is a comprehensive AI-powered development workspace. It acts as an on-demand AI pair programmer and a sovereign partner that inhabits your technical philosophy to move projects from raw intent to production-ready reality.
The Problem: Most AI coding tools force you into a fragmented "juggling act". You suffer from "context decay" where finer details are lost over time , and you are forced to manually copy-paste context or re-explain your architecture over and over. Worse, when an AI fails, it doesn't learn, frequently repeating the same generic mistakes during long sessions.
The Solution: We built an architecture that treats Human Context as a Sovereign Asset. BobCA utilizes a User Proxy powered by the Frank Engine, which clones your engineering philosophy so it makes decisions exactly how you would. Instead of a passive chat, you get an active workforce that executes surgical repairs, tests ideas in an isolated sandbox, and automatically logs its failures so it never repeats them.
What makes it different: BobCA moves beyond simple tools into the realm of Sovereign Autonomy. It provides an environment where a creator can manifest a digital representative that possesses their actual professional DNA.
Key Features:
User Proxy & Frank Engine: A dynamically stateful ambassador that carries your logical and technical DNA to negotiate with other agents on your behalf.
Linked Conversations: An Ancestry Map that shows how ideas branch, allowing you to fork contexts and seamlessly carry over passalong logic without starting from zero.
The Deep Dive Sandbox: An isolated laboratory for complex problem-solving. It uses a Project Wisdom Ledger to record "Negative Constraints" so the AI never repeats a failed approach.
Workshop Terminal: A proactive Command Center featuring an Autonomy HUD (Demon Mode) that provides a live scorecard of bugs fixed and features added in real-time.
MiniBob (Junior Apprentice): A specialized agent that handles granular, repetitive tasks in the background while you focus on high-level strategy.
Who It's For: Visionary leaders and developers looking to eliminate decision fatigue and scale their individual talent to the level of a high-performance workforce.
What We'd Love From You: Drop into Bob's Workshop, fork your first conversation, and trigger Demon Mode. We’ll be in the comments all day—what's the first autonomous repair your User Proxy knocked out for you?
@kemone_phillips Congrats on the launch, the “negative constraints” idea is interesting. I’m trying to understand the mechanics a bit more: how do you preserve statefulness across sessions and tools? Is it storing explicit rules/preferences, reading repo history, maintaining a project memory layer, or something else?
BobCA
@zolani_matebese We worked for the past year to build our Frank Engine. This is what powers the dynamic nature of context awareness across sessions for Bob and the other agents. Your preferences both technically in terms of preferred tools, architectural preferences, tech stack expectations, ui design preferences along with personally in terms of work speed, communication structure, and goal alignment is learned over time. Our engine has no limit on what he learns as we developed a new dual system memory to work more human like in preserving short/long-term memories (unlimited by the way) for better experience recall resolution. Additionally yes repo history, and project memorization is a part of the user/org specific learning.
Frank Engine against competitors :
Contextual Fidelity (Adoption to the user's specific technical DNA in a cold interaction) : FrankEngine [80%], Gemini 3.5 Flash [49%], Gemini 3.1 Pro [57%], Claude Opus 4.6 [63%], Claude Sonnet 4.6 [62%], GPT-5.5 [60%], gpt-5.4-mini [45%]
Benchmark included 3 test of across 1,000 prompts and simulated responses utilizing vanilla OpenAi GPT 5.5 for result analysis and Claude Opus 4.7 for additional verification
Environment: Brand new, empty conversation thread.
The “learns your preferences” promise is powerful, but I’d want the preference layer to be inspectable and reversible. For coding agents, a bad learned preference can quietly become a repeated bug.
Can BobCA show the user what it has learned, where that preference came from, and let them pin, edit, or expire it? That would make the product feel much safer for real project work.
BobCA
@studentzuo This is a GREAT idea. BobCA does currently show user what was learned and you can delete the restraints. However I never thought about allowing users to edit them. Pin it, or set expiration. I would love to connect and talk more about this with you. PM?
The negative constraints idea is the part that stood out to me. A lot of coding assistants are useful at first but in longer sessions they can repeat the same failed approach or lose the original reasoning. If BobCA can remember what did not work and use that context later, it could make longer projects easier to manage. How do you decide which mistakes should be saved as negative constraints?
BobCA
@busra_seker1 So negative constraints are determined by your own preferences. We spent the last year alongside partners at Nvidia and Google building out our "Frank Engine"...it's our proprietary personalization engine that was trained on user preferences, internal/external relationship matrixes, and duel memory system. NO RAG to remember all experiences. This helps our system understand you as an engineer from personal workflow style to technical habits. This self learning system acts as a healer overtime. In other words, your input with every conversation helps your version of Bob decide autonomously what's a negative constraint. Or you can always self audit.