David Smith

This video is gold if you are building agents 🙌 This guy just built out an AI Agent to build a team of AI Agents in less than 20’ You just talk to it in plain English, and within seconds, it spins up a brand-new, fully functional n8n workflow tailored to your request. It gives you a direct link to the workflow in your own n8n instance and provides a step-by-step setup guide so you can understand exactly how it works. Whether you're a beginner or an automation pro, this is the fastest way to start building custom AI agents without touching a single line of code. Here's what you can learn from this video ✔️Live Demo: Slack Agent ✔️n8n Developer Agent Breakdown ✔️Builder Agent System Prompt ✔️Claude Opus 4 Thinking ✔️Create n8n Workflow Node ✔️Some More AI Generated Examples ✔️Live Demo #2: Personalized Email Workflow ✔️My Mindset While Building This ✔️Live Demo #3: Outreach Research Workflow ✔️STEAL THIS FOR FREE ✔️Pricing Breakdown ✔️Final Thoughts I’m a big fan of n8n and this video proves you don’t need to be super technical to really master AI agents. So I really encourage you to start playing with it! Link to the video in the comments! #ai #genai #agents

536

63 Comments

The two best SaaS positioners in the game are April Dunford & Anthony Pierri. Here's their entire philosophy distilled into 10 actionable steps. Positioning will become the most important early differentiator for SaaS cos. Here is everything you need to know about their philosophies in 10 chronological bullets with clear action steps and “did I do it right?” gut checks if you read this, imo, you will know everything they have to teach about positioning: Surface the real alternatives: Action → Ask five recent buyers: “If we didn’t exist, what would you use?” List every answer (including DIY hacks and ‘do nothing’). Gut check → Your list contains at least one spreadsheet/workaround and one direct competitor. Isolate your unique attributes: Action → Circle the features or capabilities no alternative offers together the way you do. Gut check → You can count these on one hand and explain each in a single sentence. Translate attributes into value: Action → For each unique attribute, finish the sentence: “Which means the customer can now ___, and that’s worth ___.” Gut check → The blank is an outcome a CFO would care about (revenue up, cost down, risk removed). Spot the customers who feel that value hardest: Action → Rank segments by (pain intensity × willingness to pay). Pick the top one. Gut check → You can name your ICP in 12 words or fewer: “Seed–Series A PLG SaaS, 10–50 headcount, no in-house RevOps.” Choose the market frame that flatters you: Action → Decide whether you win as a new category, a sub-segment, or the best-of-breed alternative; write one headline for each and pick the clearest. Gut check → A stranger can read the chosen headline and immediately name your closest competitor set. Craft the one-line positioning statement: Action → Fill the template: “[Product] is a [category] for [ICP] that [main value] by [unique mechanism].” Gut check → It fits in a tweet without jargon and your team can recite it verbatim. Write a hero section that passes the “what/for-whom” test: Action → First line states what it is; second line states who it’s for and why they care. Gut check → Show the page (no scrolling) to someone outside tech; they can explain your product back to you in 15 seconds. Back the claim with proof, not hype: Action → Add one killer metric (“cuts onboarding time 43 %“) or social-proof logo inside the first viewport. Gut check → You quoted an actual number or recognizable brand—no adjectives needed. Align every feature and piece of copy to this position: Action → Review roadmap and marketing assets; kill or rewrite anything that doesn’t reinforce the statement from Step 6. Gut check → You can draw a straight line from each feature to the ICP’s top pain. Re-validate quarterly: Action → Repeat the “alternatives” interviews every 90 days; if answers shift, loop back to Step 2. Gut check → You’ve updated (or reaffirmed) your positioning at least twice in the last 12 months.

212

19 Comments

Here at Spark, I talked about how I believe the emergence of AI agents represents the most dramatic manifestation of the skill separation phenomenon that is and will continue driving the labor market inversion. AI agents will continue to take on tasks, make decisions and reform jobs, as a result, employees need to be upskilled and the agents themselves need to stay compliant and function as you would expect your best and most effective employees.   Which is why I’m very excited about our new partnership with Salesforce to bring agent actions to Agentforce. By bridging agent-to-agent collaboration, we’re ensuring every AI agent is trained, certified, and compliant before touching critical data and interacting with an enterprise's workforce, allowing organizations to significantly scale agent effectiveness, expanding where and how these agents can be deployed, further supporting workforce upskilling, reskilling and readiness.   https://lnkd.in/gA2bCZEJ #CornerstoneSpark25

289

7 Comments

Most founders avoid government markets. Too complex, too slow, and too much red tape. But that complexity was the opportunity. Ion-Alexandru S. noticed contractors drowning in 100+ page RFPs. He and his co-founder built GovCon in a Box to cut through the noise. What started as a chat app became a vertical SaaS that: > Solved a painful, daily problem > Turned free users into loyal paying customers > Scaled through word of mouth When the time was right, he listed on acquire.com. Clear records, steady revenue, and a simple story. Buyers moved fast. Within weeks, he had a life changing exit. Lesson for founders: > Complexity hides opportunity. > Customer feedback should shape your roadmap. > Clear, boring SaaS beats flashy features every time. Ion is now onto his next venture, seoforge.ai. Full story + podcast here: https://lnkd.in/gh9QQ-9r

29

8 Comments

Substack just rug pulled all of their customers. The creator economy’s promise is taking power from institutions and giving it to individuals... but Substack has now become the institution. They're extracting value from creators and trying to lock them into their platform for good. Last week they announced that all writers are required to offer Apple's in-app purchase (IAP) payment option, which does two things: 1) Increases the cost to readers by 30% 2) Locks writers into the platform long-term Ceding the billing relationship and platform lock-in are both suicidal for creators. This is about writer independence and creators' freedom to choose how to build their business. I explain why this is obviously terrible for everyone involved (not named Substack of Apple) in my latest newsletter: https://lnkd.in/emW3U-Se

238

85 Comments

One thing I’ve noticed with Push.fm lately ... A lot of artists are actually making real money using Push Paylinks. Not streams. Not algorithm luck. Direct-to-fan payments. That’s a big shift. Instead of hoping platforms pay out fractions of pennies, artists are selling directly — beats, downloads, exclusives, merch, access — and keeping control of the relationship. This is where things get interesting. Tools like Push.fm aren’t just marketing links anymore — they’re becoming revenue infrastructure for creators. And for many artists, that’s starting to matter more than playlists ever did.

22

3 Comments

​Usability isn't one-size-fits-all. Explore how cognitive accessibility can transform product development with Fable's co-founders. https://buff.ly/TKBS9gC #Usability Image: Five people gathered around a conference table in a modern office, engaged in conversation and using laptops. The group is diverse in age, gender, and style, with expressions suggesting engagement and collaboration.

22

AI web agents aren't nearly as capable as we think. Are we falling into an "Illusion of Progress?” Last week I wrote about the Agentic Web and mentioned the timeline would likely be more gradual than headlines suggest. Then I stumbled across this Ohio State research paper that perfectly explains why. Researchers tested six frontier web agents - including Claude Computer Use, OpenAI's Operator, and others - on realistic web tasks. The results were sobering. Most agents succeeded on only 30% of tasks. Even OpenAI's Operator, the best performer, managed just 61%. Compare this to the ~90% success rates many companies have been reporting. Here's what's really interesting about the discrepancy. The researchers found that existing benchmarks were fundamentally flawed. Many tasks could be solved with simple Google searches - their basic search agent achieved 51% success on WebVoyager, the most widely used benchmark. When they created a more realistic evaluation with 300 diverse tasks across 136 websites, performance dropped dramatically. Their search agent could only solve 22% of these realistic scenarios. The technical challenges are deeper than I initially thought: → Agents struggle most with filtering and sorting (57% of failures) → They frequently miss numerical constraints and temporal requirements → Performance drops 31% from easy to medium tasks, another 15% to hard tasks → Many rely too heavily on keyword search instead of structured navigation This connects directly to the coordination nightmare I mentioned for multi-agent systems. If individual agents can only handle 30% of single-website tasks reliably, orchestrating them across platforms becomes exponentially more complex. The infrastructure requirements I discussed - memory management, real-time data streaming, state consistency - become even more critical when your base components are this unreliable. What struck me most was the evaluation problem itself. The field has been measuring progress against benchmarks that don't reflect real-world complexity. We've been optimizing for tests, not for actual user scenarios. This reinforces my belief that business considerations should drive adoption over engineering hype. The agentic future is still coming, but we're clearly in the early experimental phase rather than the deployment-ready stage many assume. The researchers created better evaluation methods that achieve 85% agreement with human judgment. Tools like this will be crucial for honest progress measurement as the field matures. Are we seeing similar evaluation gaps in other AI capabilities we're excited about? #agenticweb #aiagents #wetechiteasy

16

Trade and building contractors waste a ton of time duplicating information across emails, job management platforms, and accounting software. We've expanded our ServiceM8 toolkit, giving Tradestack's customers a vast action space for not only the grunt work of job management, but getting the best out of this powerful system and seemless connections with email providers and accounting software, specifically Xero. Business owners can now deligate dozens of dynamic, business-specific workflows to Tradestack's agents. #jobmanagement #contractors #stopmanuelentry

25

2 Comments

What does AI actually look like in everyday engineering work? Amanda Bizzinotto from OmbuLabs.ai put it simply on The Ruby AI Podcast: ➤ “One of the most common questions we get is: How can AI make Rails upgrades easier?” When you maintain software that powers real businesses, upgrades aren’t glamorous, but they matter. AI isn’t replacing that work. It’s helping smart teams do it faster, safer, and with fewer blockers. If you’ve ever faced the do we really have to upgrade this? moment, this one’s worth a listen. 🎧 Full episode link in the comments.

8

8 Comments

The rate of AI adoption is dwarfing that of previous transformational technologies. In the next several years, AI will eat SAAS in the same way that SAAS ate on-premise software. And this is good news for nonprofits. SAAS isn't the perfect tool for resource-starved nonprofits. It needs to create work ("users") in order to add value. Agentic AI is different. Agentic AI just does the work. Want to build grant capability? Today - hire grant writer, buy grant management software, learn software, enter data, follow workflow, etc. Tomorrow - "hire" agent AI grant writer. Let it evaluate thousands of grants based on funder appetite and nonprofit capability. Write the grant. Ask for your review. Agentic AI is going to transform philanthropy. Want to hear about what we're building? DM me and you can join our private beta.

123

7 Comments