Joel Reed

CEO at OpenArc. I tweet about leadership, software development, and IT staffing. Love trail running, playing the guitar, friends, and family.

[Joel Reed @joelwreed]

Pleasantly surprised with the clean UI and seamlessly integrated AI features in @Filo_Mail. The app is in beta, and there's work to be done with resizing column sizes, message reading font size control, the archive shortcut working at all times (focus issue?), etc, but the integrated TODO management capabilities, AI summaries, filters, TODO detection, etc, are all really well done.

[Joel Reed @joelwreed]

@AmpCode can we get a setting for a default thread sharing mode?

[Joel Reed @joelwreed]

Testing and validating code will become increasingly critical as AI tools are used more frequently to implement features and make code changes.

[Joel Reed @joelwreed]

@tlamarca He who restrains his words has knowledge, And he who has a cool spirit is a man of understanding.

[Joel Reed @joelwreed]

All frontier models have a knowledge cutoff date. @Context7AI MCP gives you up to date docs & code for almost 15,000 open source libraries.

[Joel Reed @joelwreed]

Very sad day at @OpenArcNet yesterday, a dear friend and fellow employee passed away over the weekend. We spent some time sharing our memories of David, but today's not the same without him.

[Joel Reed @joelwreed]

4/4 - Dynamic Security and Permissions: In environments where access to intellectual property is restricted and permissions change over time (e.g., enterprise knowledge bases or collaborative workspaces), agentic search can adapt in real-time to ensure only authorized data is accessed and used, whereas traditional RAG's static index may not reflect these changes promptly.

[Joel Reed @joelwreed]

1/4 Where Agentic RAG/search outperform traditional Retrieval-Augmented Generation (RAG)?

[Joel Reed @joelwreed]

3/4 - Complex Logical Connections: When queries require synthesizing information from multiple sources or involve multi-step reasoning (e.g., legal analysis or scientific research), agentic search's ability to reason over retrieved data provides more accurate and comprehensive answers than traditional RAG, which may only retrieve relevant documents without understanding their relationships.

[Joel Reed @joelwreed]

How do we ensure truth in a world where AI generates content? Discover how NFTs and blockchain are paving the way for verifiable authenticity. Great conversation with @MarkCiapka.

[Joel Reed @joelwreed]

Wyatt, a 7 year old penguin, stole the show at last night's #EO Pittsburgh winter party at the National Aviary.

[Joel Reed @joelwreed]

3 takeaways from #EOPittsburgh's mid year summit:

1. Our 1st priority is always to make sure every member entrepreneur has a great forum experience, and can take part in great chapter events. Operational/delivery excellence must be a core competency.

2. Our forums are great places to make deep connections with other entrepreneurs. We should encourage building deep connections with members outside your forum as well.

3. Many board members have served for 5+ years. It's time to leverage new leadership within our chapter!

Many thanks to @prosphire for hosting our summit at their beautiful @Pittsburgh HQ, and to @cmiladinovich for his many years of service as our finance chair!

[Joel Reed @joelwreed]

Anyone leading a company all-hands in January to refocus the team on the mission, values, and specific plans for 2025? Any pointers on what works best?

I like to step back at least once a year and refocus everyone, including myself on what we are all about & why we are all about it.

Tweeting this is making me think I need to take the marketing section of my presentation up a level!

[Joel Reed @joelwreed]

@SaraMakin Running or playing guitar!

[Joel Reed @joelwreed]

Two missing @beeper features, for an otherwise awesome product.

1) filter slack messages to DM's only, no channels (or let me pick which channels)
2) google chat integration

[Joel Reed @joelwreed]

My favorite quotes from last week's readings on AI:

1) @jeremyphoward - "For those that hope (or worry) that LLMs will do breakthrough scientific research, I've got good (or bad) news: LLMs are particularly, exceedingly, marvelously ill-suited to this task."

2) @JJitsev - "SOTA LLMs cannot perform robust basic reasoning as indicated by strong performance fluctuations across even slight variations of simple problems like AIW."

3) @deedydas - "Feature steering is like brain surgery. You can steer an LLM to think how you want to. Want a model to not say "unsafe" things? Clamp those features. Want it to not exhibit gender bias? Clamp it. Want it to not generate code with security vulnerabilities? Clamp!"

4) @rohanpaul_ai - "A key to making your LLMs work better: just throw everything into the context window. For many datasets, for most of the time, the long-context ICL (in-context learning) outperforms finetuning on the model. "

#AI

[Joel Reed @joelwreed]

@EntrepreneurOrg Entrepreneurship is not for the faint of heart, but as @JordanGillissie wrote, having the "unwavering support of my wife and team" goes along way towards giving entrepreneurs the strength to see things thru. Having other entrepreneurs right there beside you doesn't hurt either!

[Joel Reed @joelwreed]

@MatthewPollard_ We've been organizing "power hours" where @OpenArcNet leaders come together, share our current important but not urgent individual goals, spend an hour moving them forward, and then telling each other what we did. Motivational and productive time.