Paddy Stobbs

Remains one of the most magical new product experiences I've had in years. Both on the candidate and company side.

@itsMeBennyB Heavy ➕

Just one more, humble request... please let me swipe left and right through my profiles. 🙏

It's here. And it's glorious. Long live the sidebar. Long live Dia.

@paulg @StatisticUrban Also, once growth stops, quality of talent that field can attract drops.

@StatisticUrban Music was a growth industry till about 1995 and movies till about 2002. The things made in a growth industry are usually better. Once growth stops, people get conservative.

I’m usually strongly opposed to the “things used to be better” schtick, but it really does seem like movies and music used to be better. Even correcting for survivorship bias, they were just higher average quality. We still get the occasional gem, but nothing like it was.

"Anyone can build anything, technical skill doesn't matter, taste is the moat" This is a fashionable notion that is stupidly reductive. If you've worked on even a moderately complex product, you know that *taste* is more than picking the right feature idea or UX. The finest engineers I've worked with share very few qualities in common, but exceptional taste is one of them. Because technical skill is itself largely an expression of taste once you move beyond junior-level competence. By taste, I mean: - A profound understanding of the domain and how to model it elegantly in code - An intuitive sense for identifying the right abstractions - Empathy for those who will maintain and extend the system Maybe this kind of taste seems trivial in our current era of vibe coding and Lovable demos. I'd argue the opposite. Coding agents enable us to generate huge volumes of code at lightning speed. The taste of those who can properly understand this code becomes the only safeguard preventing your product's foundations from rotting out from underneath you. Taste isn't merely aesthetic preference, it's a fundamental engineering discipline that matters now more than ever.

What are the best startup hiring platforms or talent networks (for commercial or engineering/design)? Things like: @wttj @hackajobHQ Jack&Jill

I watch a lot of Air Crash Investigation, which means I've seen the same pattern play out almost every time: most crashes happen when pilots take over from autopilot, not when the automation fails. Now I'm seeing the same pattern in AI-assisted coding. Aviation is incredibly safe BECAUSE of automation. But there's a dark side. When pilots who rarely hand-fly need to suddenly take control, disaster strikes. They've lost their edge. They trust the system too much. Software development is speedrunning this exact problem. Devs are letting AI write more code every day. But what happens when they need to debug something they didn't write and barely understand? In aviation: Accidents spike when pilots override autopilot. In software: Bugs emerge in the gaps between AI-generated code and human understanding We're creating the same dangerous knowledge gap. I honestly worry a bit less about the developers. You write tests, the app isn't super critical, whatever. However I think users, particularly in critical industries have no idea their apps are built on this hybrid human-AI foundation. They will think it's all carefully crafted by humans, like passengers assuming the pilot is actually flying the plane. The risk isn't that AI fails constantly. It's that when it does fail, nobody's prepared. We're building a world where neither developers nor users understand what happens when the automation breaks.

@chalmermagne The government's position that upending copyright law is somehow beneficial to writers/artists/musicians.

can anyone think of good e.g.s of 'noble lies' in policymaking - assumptions which we pretend are true (for convenience, simplicity, public reassurance or w/e) long after all the evidence has started to point in a different direction?

This 👇. Garbage in, garbage out. The great LLMs-will-remain-horrendously-unhelpful-in-fields-where-the-public-data-sucks awakening is upon us.

Dig into all 14 of our Lite Categories here 👉 go.stackfix.com/litecat

NEW - Lite Categories Excited to announce that 14 new Lite Categories are now live on Stackfix 🎬 Including.... - Customer Success (feat. @planhat , @vitally_io , @Totango ) - Business Intelligence (feat. @omni , @lightdash_devs , @metabase ) - Business Phone Systems (feat. @JoinNooks , @aircall , @justcall_io ) - Employee Benefits (feat. @joinHappl , @_ThanksBen , @getbenepass ) - Treasury Management (feat. Round, @join_arc , @rhobusiness ) These are categories that we’re still testing - but they contain our early recommendations. Deeper testing and full analysis coming soon. Check them out at the link👇 And if there's a product you think deserves a spot in our full testing, we'd love to hear about it!

Check out our full deep dive - with granular ratings and analysis 👉go.stackfix.com/empher

5 products in 1. Streamlining is a massive theme in software right now, and one product that leans into this is @EmploymentHero - an HR platform that offers a smorgasbord of people-related functionality. Core HR, Payroll, ATS, Benefits and Global Hiring all in a single tool. But is it any good? We took it for a spin. Check out Kwok Cheung's video overview to get a snapshot of the pros and cons, and our full analysis via the link in the comments👇

@StephNass Cheers a lot Steph - and likewise on OpenVC - super useful!

@PaddyStobbs Thanks Paddy! I've been looking for sthg like Stackfix - even thought of building it myself. Great execution, too!

NEW - Product Analytics 📈 One of the first decisions for any software company is - which product analytics software should we use? Mixpanel? Amplitude? PostHog? Hotjar? Or just basic Google Analytics. The truth is that each of these are built for slightly different use-cases. And the right choice depends on what you need to track + and how much complexity you're willing to handle. So we've tested and compared the leading products. So you don't have to. Shout with recs of what other analytics tools we should prioritise adding!

I've been thinking about how to explain the distinction between an MCP server and an API, and landed on an analogy I like: the MCP server as an encouraging parent. Basically imagine a parent & child + Lego bricks. The LLM is something like that child - energetic, occasionally confused, but nascently agentic. The role of the parent (MPC server) is to provide resources the child can't reach, help snap pieces together, and offer nudges toward some goal. The Lego bricks are just the collection of resources on the web exposing some API - your filesystem, the weather API etc. So it's clear that the MCP server is really acting to bring those resources within reach of LLM interaction.

👉 Fuller review here - go.stackfix.com/closecrm

Tested a lot of CRMs - and @Close is one of the strongest options out there for early-stage teams focused on high-volume outbound sales. It particularly nails cold calling - with some of the best phone features on the market. My cliff notes 👇 (N.B. Not an ad – just love testing products) 🪧 Name: Close 📆 Last Tested: Apr 2025 ❓ What: Sales CRM 👥 Used by: @Hipcamp , @branch , @MakeSpace etc 🏙️ Founded: 2013, HQ in SF 🇺🇸 👉 What does Close do? CRM that is particularly geared towards sales outreach by letting you: - Make cold calls at scale with power and predictive dialers - Automate multi-channel sequences (email, SMS, calls) - Segment leads using advanced Smart Views and custom pipelines ---------------------- ➞ My take: ➕ Best-in-class for cold calling & SMS – Close is exceptional for high-volume cold calling. Its power dialer, predictive routing, AI transcriptions, and call coaching tools make it a standout for teams making 50–500+ calls a day. ➕ Flexible and fast – The interface is quick, the layout is intuitive, and Smart Views let you jump straight into filtered lead lists. Custom objects and pipeline flexibility mean it can adapt to complex workflows without slowing you down. ➕ Multi-channel sequences – Close makes it easy to automate outreach across calls, texts, and emails – complete with personalization, delays, and A/B testing. The analytics help you see what’s working, and adjust fast. The Reality Check? ▬ Limited workflow automation – Beyond sequences and basic status updates, Close doesn’t offer much in the way of process automation. There’s no if-this-then-that logic or triggers based on deeper events. ▬ No lead scoring – For a sales-focused CRM, the lack of lead scoring is surprising. If you’re working large volumes of leads, it makes prioritization harder than it should be. ▬ No marketing features – Close doesn’t offer marketing automation, templates, or a proper drag-and-drop email builder. If you need lifecycle email or newsletter-style campaigns, you’ll need to integrate another tool. ---------------------- Overall? If you're an outbound sales teams doing high-volume, multi-touch outreach – especially if calling is a big part of your workflow - then Close is a very strong option. But if you’re looking for deep integrations, marketing features, or advanced automation, there are more appropriate tools. Much more to it though – full deep dive below 👇

In 1764 James Hargreaves invented the Spinning Jenny - and fabric production went from slow & expensive to fast & cheap. = The number of producers exploded = Differentiation got much harder = Trusted brands rose (Hermès - 1837, Levi's - 1853, Louis Vuitton - 1854) Feels like a useful analogy for what might be about to happen to software. When the means of production get commoditised and the market is flooded, buyers cry out for trusted brands. To signal quality and to get confident that the product they’re shelling out on won’t break/have a data leak at an embarrassing moment. And I reckon - given the coming AI-powered explosion of low-quality products, and how mission-critical software is for many businesses - that we’re in the infancy of trusted software brands right now. So if you're vibe coding without thinking of the brand you're building, you're the slop-shop tailor - whilst Levi Strauss is stitching the 501.

ChatGPT is trained on a lot of garbage. - Fake/incentivized reviews (Amazon, G2, Trustpilot) - Reddit comments - Quora answers - Propaganda from the far ends of the political spectrum - Tabloid journalism etc etc As most people know, this means that ChatGPT outputs a lot of garbage (factual errors, bad advice etc). But because LLMs are genuinely impressive in *certain* fields (where there's abundant, high-quality training data - e.g. coding, or how to boil an egg) it feels like we're lazily expecting the big name LLMs to get really good at *all* fields. They just won't. In domains where the publicly-available training data remains garbage (e.g. private company data, or product evaluation, or why your girlfriend really dumped you) then no breakthroughs at the model layer are going to help. What you need is abundant, high-quality training data. And, with most of the low-hanging data fruit picked (or stolen), then the next wave of breakthroughs will come from people who own or generate data that no one else has. That's what we're doing.

Slop blog posts Slop music Slop games Slop companies (”Hey Lovable, make me a dating app”). = Tsunami of slop incoming. Never been more alpha in building trusted brands.

@saranormous "They often have momentum in the recruiting process" - one of the truest things I've read in a long time.