Mo ShaRaf (sharaf.eth)

Knowledge seeker | Strategist | Enterprise Architect | Tech Consultant | AI & Web3 Researcher | Mentor & OSS advocate | Looking to bring a positive impact 😊…

[Mo ShaRaf (sharaf.eth) @MSharafH]

GenAI’s impact reaches far beyond coding. The biggest performance uplift is happening in testing, deployment and maintenance - the stages that determine day-to-day reliability.

As pressure grows on teams to deliver more with less, these downstream efficiencies are becoming essential.

This is a space worth exploring for any organisation looking to strengthen resilience while accelerating delivery.

[PwC Middle East @PwC_Middle_East]

From faster releases to higher quality code, agentic AI is reshaping software delivery.

84% of AI adopters report faster delivery and improved code quality.
Pioneer teams embedding agentic workflows across the SDLC are averaging more than 70 releases per year.

If you would like to read more, please click here. https://t.co/p8iwu3eUEE

#AgenticAI #DigitalTransformation #SoftwareDelivery

[Mo ShaRaf (sharaf.eth) @MSharafH]

Who would think back in 2020 @obsdmd will be the ultimate personal knowledge base for AI agents.
Glad that I am an obsidian user since then :).
It’s much easier to pull adaptive knowledge into my agents…

[Mo ShaRaf (sharaf.eth) @MSharafH]

@RoundtableSpace What an idea……

[Eliza Kosoy @ElizaKosoy]

Excited to share that I've left Character AI to start a new company.

I'm building a frontier AI lab and public benefit corp, bringing together work from my PhD at UC Berkeley, Google DeepMind, and Character AI. It's the culmination of my life's work at the intersection of youth empowerment and AI.

We're currently in stealth and hiring our founding team, including a founding engineer and designer.

If this resonates, or if someone exceptional comes to mind, please reach out or send them my way.

visit https://t.co/xEW3ZIprVr for more roles and info on how to apply

friends plz retweet for viz

[Dario Amodei @DarioAmodei]

Today I'm publishing a new essay, Policy on the AI Exponential. AI is progressing extremely fast—much faster than the policy process was built to handle. The essay lays out where I think the technology is now, and the action needed to close the gap: https://t.co/Lh6PWae178

[Mo ShaRaf (sharaf.eth) @MSharafH]

@MatthewBerman But if we used a local model that will have more frequent running loops to increase output quakity through iterations, the limitation won’t be tokens… it would be time to market

[Mo ShaRaf (sharaf.eth) @MSharafH]

@MatthewBerman @steipete +1

[Mo ShaRaf (sharaf.eth) @MSharafH]

@steipete Indeed, but you need a well-articulated defined goal :)

[Mo ShaRaf (sharaf.eth) @MSharafH]

What is First Principles Thinking? https://t.co/bM3vMVwjRH
A good enriching read.

[Mo ShaRaf (sharaf.eth) @MSharafH]

@garrytan I didn’t know this already exist. What a game changer for Agents ecosystem

[Mo ShaRaf (sharaf.eth) @MSharafH]

For a while AI giant labs are eating startups on breakfast, everything AI native or based idea or product becomes the lab priority which they can easily develop, maintain, and scale with their current offering. Making it harder for others to compete
Few startups will remain solid and competitive due to their strong and unique harness like Replit.

[Mo ShaRaf (sharaf.eth) @MSharafH]

@karpathy All the best, this is great news indeed.
I am a huge believer in Anthropic unique mission and in your capabilities will blend their in a unique way too.

[Mo ShaRaf (sharaf.eth) @MSharafH]

@steipete Perfect depiction and use or your in-house built stack
I will be building similar dev center but with local open source model that i can afford :)

[Mo ShaRaf (sharaf.eth) @MSharafH]

@MatthewBerman There is a real barrier as well, you mentioned it in the beginning, US stock market is build around US tech companies achievements which something great, but also hinder the open source mind set, otherwise this will cause a huge damage for the market

[Mo ShaRaf (sharaf.eth) @MSharafH]

@MatthewBerman Build your own tool, suggest checking Garry GStack

[Mo ShaRaf (sharaf.eth) @MSharafH]

@ycombinator @garrytan Hats-off Garry, indeed a state of art, eager to try it tonight, will make our life much eaiser and save some tokens over the way

[Mo ShaRaf (sharaf.eth) @MSharafH]

- Rich knowledge dense for normal AI applications or prompts will make your AI responses weak, you will struggle with converting it to something really useful
- While having an isolated well curated small knowledge points will make your AI applications or prompts more efficient specially as LLM models are improving significantly overtime

This is my take for enterprise agentic workflows and for personal AI knowledge portable basis.

Trick is in the index to your own universe

[Mo ShaRaf (sharaf.eth) @MSharafH]

@amasad Great paper, as the AI coding agents are augmenting code / system generation, security harness is also required mixing both Human expertise and AI patterns classifier capabilities

[Mo ShaRaf (sharaf.eth) @MSharafH]

@MatthewBerman He is flashing the money right now,
🤣🤣🤣🤣🤣🤣 hilarious

[Mo ShaRaf (sharaf.eth) @MSharafH]

@garrytan Nice criteria, while I would say, OpenClaw is your ultimate gateway to impersonate anything you would imagine let it be a bus in the beginning till reaching your star trek space machine

[Mo ShaRaf (sharaf.eth) @MSharafH]

@MatthewBerman Like JAVA old people, you need a harness optimizer equivalent to garbage collector in JAVA for optimizing memory
The labs are investing and focusing on models improvement not harness optimization ( actually it’s better for their revenue this way)