Readit NewsBuilt Cognitora.dev after getting frustrated with existing AI coding environments being either too limited, too heavy or too expensive. We're using Kata containers with the flexibility to run on either Firecracker or Cloud Hypervisor - letting us optimize for the best performance per workload.
Key differentiators from e2b/Daytona:
- True isolation with hardware-level security (not just Docker) - Sub-3s cold starts vs 10-30s elsewhere - Kata flexibility: Firecracker for ultra-fast startup, Cloud HV for heavier workloads - Built-in AI context preservation across sessions - Works with any LLM/model (not locked to specific providers)
The sandbox handles everything from simple scripts to full deployments. We're seeing devs use it for prototyping, code review with AI, and even production debugging in isolated environments.
Still early but would love feedback from the community.
Website: cognitora.dev
Docs: https://www.cognitora.dev/docs/getting-started/examples
AI tech support, AI summarizers, try them! Are you actually impressed with them? Be honest. I find them infuriating, stupid, and pointless. You can't trust what they say.
You know how people feel about used car salesmen? They also feel that way about AI sales reps.
I'd like you to sit next to me the next time I try to coax actual details and puzzle through contradicting statements from a tech support bot. I will periodically turn to you with a pleading look on my face, as if to say "Did you see that? And you think that's okay?" and then I will open my mouth and say the same words that my face shouted.
Undortunately, it's getting to the point that the moment I suspect I'm dealing with AI I am out.
You should pre-record your demos and provide documentation and contact info. Absolutely nobody is impressed by your fucking bot.
That's why in the beginning of the call it discloses it's an AI. Not all people are cool with it, but for most people who are not into tech, it's actually a fun experience I would argue
I mainly use Google Meet's API (hosting the meeting), OpenAI GPT v4o (llm), 11labs (for voice), deepgram (for the transcription of the meeting) and Twilio (to connect to the meeting).
Of course, it's early days (I built this yesterday) but so far, the results looks promising.
Unlike pre-recorded demo videos, AI agents can screenshare and demo the product while also responding in real time to questions and handle objections as they arise.
What do you think?
Thank you for posting my thread!
Feel free to ask questions on this =)