Syed Muzayan Mehmud
184 posts
Syed Muzayan Mehmud
@syedmuzayan
Making LLMs actually useful for work → building 🔍 https://t.co/Dvf5KGRYE1 · 🧠 https://t.co/lR0LT7183J Ex-Principal Consulting Actuary
USA Katılım Temmuz 2016
20 Takip Edilen9 Takipçiler
so yeah he obviously needs to shut up with this talking point
but also why is this literally the only thing he ever says. like every few weeks he just says the same thing again with new made up numbers
if only claude could make new material
TFTC@TFTC21
Anthropic CEO Dario Amodei: “50% of all tech jobs, entry-level lawyers, consultants, and finance professionals will be completely wiped out within 1–5 years.”
English
Why isn't USA in the pick list for region when trying to download claude for excel from msft site? The first entry is Antigua, and it's a required field. @claudeai @Microsoft
English
@MSiekmoeller I like the breaks. I also sometimes try a different model, that’s fun too.
English
@PeterDiamandis I don’t see how saas margins don’t compress. At least.
English
Did you see "Claude Design"? TOTAL DISRUPTION… Canva and Figma are freaking out… As will ALL other verticals. The LLM companies will disrupt every specialized vertical use case in order to maximize their revenue, to build out their data centers and close their business case. Fireworks are just beginning!
English
@sherifgjini Dogfooding my product. For example, security scanning my repo, marketing copy, and of course - building the product that makes all that work. It’s an ouraboros
English
Codex way better than Claude Code for me now. Plus I don't run out of tokens before lunch
English
@thepericulum You have to pay api cost if you want to use sonnet or opus in cursor, right? I agree on ux. Model mileage varies with tasking.
English
Cursor is so much better than Claude Code
Output quality and UX
It’s sad most people don’t see it
Theo - t3.gg@theo
"Clearly Claude Code isn't that bad, millions of people use it!" This mindset is how we got stuck with Internet Explorer. I'm gonna keep pushing for us to do better as an industry.
English
@garrytan Markdown is great. Serving decision logic via mcp is even better
English
"It's just markdown files"
Dude anyone who believes this hasn't built agentic software yet
Cypher Samurai 🍃@Cypher_Samurai
@garrytan @zeeg maybe cause, think about it gary... maintaining a bunch of markdown files isn't real work?
English
I agree. I also think the only durable moat in the end is human expertise. And you cannot encode that for any complex process through a markdown file or a a set of docs or even a db.
I've been focusing on how to make it easy to encode expertise that can shepherd a complex agentic workflow (e.g. a consulting firm deliverable). I think the only thing that works is a certain kind of decision tree that then guide any agent you choose to work with.
English
To get LLMs to work effectively, you need to extract and document all the implicit "domain knowledge" in your business.
Once you do that, AI agents are magical.
But it also means that OpenAI/Anthropic have full access to your company's domain knowledge 😬
English
Anything that locks your context with a vendor, inference provider or a frontier model - is a bad idea.
English
FIGHT! Watch the Official Trailer for Street Fighter, hitting theaters everywhere October 16.
English
Still an early beta , but worth trying out
Tech Dev Notes@techdevnotes
Grok 4.3 can create PDF and almost any file formats like Sheets, Slides, Doc
English
@ConnerBean I wonder whether it will flatten the bell curve, or make it bi-modal
English
AI is making the average engineer worse but the best even better
English
In order for AI to be useful for real work rather than create slides, draft email, or summarize data - it needs to follow a complex methodology.
Human expertise isn't a markdown file or a prompt. We need something other than what exists today to effectively encode it for AI agents.
Which is why I built mcpforx. I pushed the complexity on this example, 24 separate skill instructions and 51 edges/decision points. Connect any AI agent (cursor, claudecode, codex etc.) and it can follow your workflow.
In this case a life insurance underwriting process that routes applications through simplified issue, full medical underwriting, or group scheme tracks, each with their own assessment logic. I'll post a link in comments, comes with sample prompts you can use to test the concept with your AI agent.
English
we had to declare bankruptcy and start over
painful decision/process and it set us back a bunch
but now we have the foundations to try and do something good
brendan@brendonovich
we've moved opencode desktop to electron. it's faster, more reliable, and will replace our tauri build soon. try it out in beta via the link below.
English
If OpenAI is the promise and MSFT the reality when it comes to enterprise. It seems the market is simultaneously bullish on the hype and bearish on the reality.
English
Excited to announce the first major product I’ve worked on at @OpenAI : Computer Use in Codex!
Our team has been hard at work bringing this magical experience to you over the last couple of weeks
Let us know how you use Computer Use!
Some of my favorite details 👇
OpenAI@OpenAI
Codex for (almost) everything. It can now use apps on your Mac, connect to more of your tools, create images, learn from previous actions, remember how you like to work, and take on ongoing and repeatable tasks.
English
I think we need a great OSS library that can auto-scan github repos and give it a check mark and update on each merge. It would be a huge services to OSS, and github should probably spearhead. Don't need mythos for it, I think even a cheaper model would be a giant step up from where we are.
English
Open source software will be many times more secure than closed source software in the new Mythos era
Peter Steinberger 🦞@steipete
they: OpenClaw is so insecure look at all these GHSAs! reality: we are just an indicator of the coming storm
English
@ycombinator @DatostApp @MaceoCk Does it accept MCP? Semantic layer is one important aspect. Another crucial one is the methodologies and expertise of the team or user. Wondering whether what I'm building plugs into this.
English
Datost (@datostapp) is an AI data analyst in Slack.
It keeps a semantic layer of your business definitions, crm, docs, and codebase so it knows what questions mean. 75.2% on the hardest public text-to-SQL benchmark, where Opus 4.6 scores 33%.
Congrats on the launch, @maceock & @jasonhywang!
ycombinator.com/launches/Pxg-d…
English