Pinned Tweet
I just open sourced my ai assistant edward! Check it out at meet-edward.com
He has memory he thinks about, the ability to evolve and some other kinda unique shit
1
34
Had a thought about what life is like when everyone has ai assistants.... (if they do...)
Does this resonate with anyone?
zyroi.com/blog/when-your-ai-…
15
I built an AI assistant that actually remembers who I am.
Not "remembers for this chat session." Remembers across every conversation. My preferences, my projects, my contacts, my schedule — everything.
It reads my messages, schedules itself to follow up, runs code, sends texts on my behalf, and learns more about me every single day.His name is Edward.
1
30
Here's what makes Edward different from every other AI wrapper:
— Long-term memory that grows over time
— Voice conversations with sub-200ms latency
— Reads and triages your iMessages autonomously
— Schedules its own reminders and follow-ups
— Runs Python, JS, SQL, and shell commands
— Integrates with Calendar, Reminders, Contacts, Mail
Early access waitlist is open → mredward.org
1
21
The wildest part is watching him connect dots I forgot about.
I mentioned a friend's birthday in passing a month ago. Last week Edward reminded me unprompted and drafted a message.
That's the difference between a chatbot and an assistant. One answers questions. The other actually pays attention.
Still unsure of how much freedom I want to give him...
14
I created a pwa to make HN easier to read. I took a ton of inspiration from @ChristianSelig apps of old…
ben4mn.github.io/HN/#/top
25
I wrote a blog about writing an app - zyroi.com/blog/app-in-a-day
Would love any feedback - trying to get better at #BuildingInPublic and participating in Claude Code convos. Thanks @AnthropicAI for the holiday usage boosts!
24
Hahaha this is what my year end review with GPT gave me... I think my team would vouch it mirrors my annoying tendencies at work
12
Fun little SaaS poc I'm toying with that lets a user create temporary bots with a dedicated RAG lookup that they can embed in Slack or Zoom (for specific knowledge or complicated change processes) and then remove when the analytics show the need has passed. #SaaS
1
52
Hey @awscloud how come when I woke up this AM I am getting "you've reached your daily usage limit please return tomorrow" errors with Kiro when I haven't used it at all in the last two days? Does everyone get that "limit reached" as an error during high load? I know its free but cmon
1
1
101
I wrote another rambling blog post - would love your thoughts and critiques
bf.weblog.lol/2025/07/the-in…
#FutureTech
1
11
What I assume everyone vibing with Kiro is getting today... good for them with the must be insane traffic (free tokens for all...)
28
The only good thing about @grok is that is shits on twitter so hard 🤣
68
I built an MCP server for Claude Desktop that lets you search for flights from Amadeus GDS...
github.com/ben4mn/amadeus-mc…
it was more for learning... I'm aware its kinda shit, open to all the feedback!
44
Benjamin retweeted
👀 Bernie Sanders on How the Democratic 'Establishment' Took Him Out of the Presidential Race
"That's what happens when you take on the establishments. What happened is we won the first three primaries, and then the establishment got very, very nervous...and they said, 'Hey, it would be a good idea if you dropped out'...So if you're asking me, are we a democracy? In one sense, we are...I use the term oligarchy. An oligarchy is a society where small numbers of very wealthy people control the economic and political life of the country. I think we are moving rapidly in that direction."
@TheoVon @BernieSanders
It finally happened! I landed at B90 and walked to B8, those are the most opposite ends of DIA! Haha thanks for the steps @united (not sure why it ends at B8) at DIA?
33
Benjamin retweeted
Spotting LLMs With Binoculars: Zero-Shot Detection of Machine-Generated Text
paper page: huggingface.co/papers/2401.1…
Detecting text generated by modern large language models is thought to be hard, as both LLMs and humans can exhibit a wide range of complex behaviors. However, we find that a score based on contrasting two closely related language models is highly accurate at separating human-generated and machine-generated text. Based on this mechanism, we propose a novel LLM detector that only requires simple calculations using a pair of pre-trained LLMs. The method, called Binoculars, achieves state-of-the-art accuracy without any training data. It is capable of spotting machine text from a range of modern LLMs without any model-specific modifications. We comprehensively evaluate Binoculars on a number of text sources and in varied situations. Over a wide range of document types, Binoculars detects over 90% of generated samples from ChatGPT (and other LLMs) at a false positive rate of 0.01%, despite not being trained on any ChatGPT data.







