Alexander Benz

@OtiumLabs @mal_shaik I think so lol

@mal_shaik @abenz_mato lol is this rage bait

i've been copying and pasting code from chatgpt but there has to be a better way how come everyone is just accepting this?

@OtiumLabs @t_blom Thanks! Let me know if you want a free demo episode or be a guest on one of our podcasts.

@abenz_mato @t_blom Probably the only comment I’ve read that makes sense. Cool biz too.

Counterintuitively, I see some startups obsessing *too much* over their first customer. Big enterprise customer ghosting you? Maybe they’re just not a good fit. Go get 10 more. Top-of-funnel solves most problems.

@samanthajeanneb The trait is real but the object matters. Obsessive about the customer's problem pulls a company forward. Obsessive about the product alone pulls the founder into a corner.

The longer I spend in SF the more I realize the #1 trait for a successful founder is being obsessive

@arian_ghashghai Best cold inbound we got for Mato came from someone who opened with 'your changelog last week is wrong about how guest pipelines compose at scale.' Walking into the weak spot bought the meeting in one paragraph.

That being said (and to manage expectations a bit), most cold outreach messages suck. Some pointers to make them suck less: > acknowledge cold messages are needle in the haystack stuff > your message should absolutely include why *you* are great for the problem the you’re solving (a remarkably high number of emails I get tell me nothing about the founder. Note that pre-seed is primarily underwriting the founder) > spell it out to the investor why you believe the should care. There are 2 solid ways of doing this: 1) most VCs are quite public with their opinions (by design), and connect what you’re doing with something that investor has gone on record caring about and/or 2) connect what you’re doing with an existing portfolio position. (however, a lot of the time I see a founder say “I saw you invested in X, so you should like my startups which is doing exactly what X is doing”. This is a bad way to do it, as most VCs are unlikely to back a competitive company (esp from cold email). Instead your message should be how your startup might be complimentary to what they’ve already deployed money in)

@KaiXCreator Half true. You need enough code to read what the agent gave you and notice where it went wrong. You don't need enough to write it yourself anymore. The bar dropped from 'can you author it' to 'can you recognize the broken thing in line 47.'

Hot take: Vibe coding only works well if you already know how to code.

@aaron_epstein Most "this time is different" arguments are right about the model layer and wrong about everything built on top of it. The startups worth watching in 2026 are shipping workflow the foundation model can't ship alone.

Every generation has a company that seems inevitable. Microsoft in the 90s. Google in the 2000s. Facebook in the 2010s. Anthropic/OpenAI now. It always feels like it's different this time. It never is. Startups always find a way.

@chhddavid The 'two founders + AI + zero employees' model is real. We're running Mato the same way. Four agents handle the podcast production pipeline. One human handles the sales calls. The constraint moved from headcount to context window.

so you're telling me Claude Code Opus 4.6 can now... - build an app - launch it - do email marketing - self-build features without any human in the loop?? it's so over... x.com/chddaniel/stat…

@pmddomingos Vibe thinking is the entire job now. The agent removed the typing. The bottleneck moved to writing the requirement specific enough that the first pass is right.

The problem with vibe coding is vibe thinking.

@sirsamjenks Standing in front of a CEO and naming the line in his P&L that I'm going to fix. The agent can rewrite the deck. The agent can't sit in the seat and look him in the eye.

do what the agents can’t do

@zuess05 Distribution. Plus an unfair process. The product is now table stakes. The thing that makes a company valuable is whether you can drag 10 customers to a 30-minute spreadsheet call and close 3 of them.

If shipping a product only takes a weekend now... what actually makes a startup valuable in 2026?

@ColinGardiner Same pattern here. My Claude Code usage dropped 60% after the first month of Mato. The 90 minutes a day I spend in it now are on the two files that actually matter. The other nine the agent owns. Steady state is small.

Getting tired of using my different agents. I enjoy my line of work because of the people interactions. My steady state usage of AI is definitely lower than I expected. Still use it a lot to be clear.

@xuster The question I wish I'd asked five years earlier. Most pivots are downstream of 'top 5' being treated as 'top 2.'

Founders often ask me what to listen for in customer calls to know if they have a hair-on-fire problem. This is the wrong question. Can you run their business? If you were CEO, would this be a top 2 priority? Because most businesses never get to #3. Rather than pitching your idea, learn the business so well you could run it yourself. You'll know exactly how big the problem is and what they'd pay. And if this is going to 100x their business and they still won't buy? You just discovered the AI version of their company.

@jasonlk True in podcasting too. Hosts who pitch us close in a week. Networks we pitch are six months in. The picker-vs-picked split happens at every layer of the deal funnel.

The prospects that pick you usually close the fastest The prospects that you pick often have the highest deal size

@lennysan @danshipper Same thing happening to podcast production. I'm spending more time in Claude Code planning episode arcs than I spend in any audio tool. The agent has become the work surface.

My biggest takeaways from @danshipper: 1. The future of work will happen inside Codex or Claude Code. Instead of putting AI into your SaaS tool, you’ll use your SaaS tools inside your favorite AI agents' in-app browser. Dan spends all his time in Codex now—writing documents, managing email, doing research, everything. He's using Google Docs, PostHog, and everything he needs within the agent's in-app browser. The agent can see what he’s doing, and has all of his context, so he and his agent collaborate quickly and super effectively. 2. Automation is a lie—every automation needs a human. Dan's company doubled in size this year despite being incredibly AI-forward. Why? Because in order to make automation work well, you need humans making sure everything keeps working. This is why benchmarks are misleading—they measure AI on problems we’ve already framed and can score, but there’s always a higher frame. 3. PMs will win the AI era. Marcus, a former PM who previously ran Axios’s writing product, joined Every after getting super AI-pilled. Now he runs their product Spiral, and ships faster than anyone on the team. He pairs technical knowledge with spiky product sense, deep user empathy, and an eye for what matters. Dan thinks any PM who gets really AI-native will be incredibly dangerous because the building is done for you—what matters is figuring out what to build and if it’s great. 4. Full-stack designers are becoming superheroes. Designers used to make beautiful interactions that engineers didn’t want to build or couldn’t execute properly. Now designers don’t need to hand things off; they can build it themselves. Designers are naturally creative people, and AI is the perfect tool for them because it lets them bring their vision to life without the traditional bottlenecks. 5. SaaS is not dead. In fact, Dan is bullish on SaaS stocks. When users bring their own AI (via Codex or Claude Code) to use SaaS products, the user—not the SaaS company—pays for tokens. This saves SaaS company’s margins. Since the agents need their own seats, Dan predicts that agents will create massive new demand for SaaS because there will be tons of agents using these products at high volume. 6. Every company will have one “super-agent” inside their Slack that every employee will use. Dan initially thought every employee would have their personal work agent, like a shadow AI org chart, but he’s completely flipped his view. He realized agents need humans who care about them. When someone gets tired of maintaining their personal agent, it becomes useless. The winning model is one forward-deployed engineer or AI-savvy person who maintains a company-wide agent (like Shopify’s River or Viktor), and then it trickles down to more specialized team agents as models improve and become less fiddly. 7. The AI job apocalypse is not happening, but you do need to evolve to stay relevant. Models make yesterday’s human competence cheap. But because everyone uses the same models, it all looks the same if you use it the default way; it becomes commoditized slop. Humans then take that frozen competence and use it to make something new and interesting for their specific situation. The key: “ride the models”—use them for everything you do, try new models when they drop, keep turning over rocks. 8. We will read way more AI-generated writing, and we will like it. Human writing is incredibly important for things that matter, but for internal docs, planning, and email, AI-generated is often better because most people are bad at writing strategy documents. 9. Build software for humans and agents to use together. The current model is building a CLI that an agent uses independently. Instead, you and your agent should be using the app together. This creates new design challenges—agents can make a billion requests in three seconds, so you need approval flows, inboxes that summarize what happened, logs, and easy rollback. 10. Forward-deployed engineers are the new most essential role. The big model companies have teams of people managing their internal agents, and those teams aren’t going away. It’s different from traditional software building, and certain engineers love it. As models get better, this role will evolve—you’ll be managing more agents doing more things.

Funny how Codex is so sure it implemented 100% of a task, upon closer review, it was only 30%. Codex has really gone downhill over the last two weeks.

The business model problem is the real blocker. Live means always-on compute, and always-on compute means you need to charge enough to justify servers running 24/7 for each customer. Most AI products can barely justify the inference cost per query. Continuous monitoring is a different cost structure entirely.

one of the most interesting things about ai products today is that almost none of them are *live*. there’s nothing running continuously, reacting to context as it changes.. maybe a scheduled digest here or a timer there, but that’s just pull dressed up as push. everything is fundamentally a vending machine where you walk up, ask, get an answer, & then leave. getting this right is obviously tricky & the business model behind must fit to justify the burn but this is where really interesting application layer problems live rn.

@tunguz The terminal-only upgrade is the wild part. Most people assume these tools need a GUI to be useful. Codex running on one machine and upgrading another over SSH is closer to how sysadmins work than how developers work. Different ceiling.

I am making Codex upgrade my very old System76 laptop running Ubuntu 16.04 with 1070 GPU. Codex is running on my Mac laptop and upgrading the Ubuntu one via terminal. If it pulls it off, I'll concede that we have AGI already.

@pierreeliottlal Same approach at Mato. We sold enterprise podcast deals with a deck before the product was done. 3 of 4 customers signed after a single 30-minute call walking through unit economics on a spreadsheet. The product got better because customers told us what to build.

Most SaaS founders fail because they scale acquisition too late. We did the opposite. We focused on distribution before product perfection. That’s how we went from $0 to 2,000+ paying customers in 9 months with 0 outside funding. Here’s the exact growth system we used at GojiberryAI: STEP 1: Validate demand before building Most founders spend months building features nobody asked for. Instead: we started selling before the product even existed. We created a simple 7-slide deck: • the problem • the workflow • the expected result Then we started outbound immediately. Using GojiberryAI + manual sourcing, we targeted high intent leads through LinkedIn and cold email. Weekly targets: • 180 LinkedIn invites • 1,000 cold emails • 4-8 demos per day That was enough to validate demand. The signal we looked for was simple: people asking: “How much does it cost?” Before the SaaS existed, we were manually selling curated high intent lead lists. Once enough companies bought repeatedly, we automated the workflow and turned it into software. The goal at this stage is not scale. It’s getting your first 100 customers as fast as possible. STEP 2: Turn customer results into growth Once customers started getting results, distribution became easier. Our biggest unlock: Reddit. Not ads. Not SEO. Not partnerships. Just giving massive value publicly. What worked: • commenting on viral posts • sharing tactical breakdowns • posting real customer wins • explaining exactly how we generated pipeline No corporate branding. No polished marketing language. Just actionable content. At the same time: we kept outbound running aggressively every single day. First hires: 3 Customer Success Managers. Retention and customer wins became growth loops. STEP 3: Build an inbound engine Once we crossed ~$25k MRR, we doubled down on LinkedIn. Every person on the team: • posted daily • handled outbound daily We only published lead magnet style content: • frameworks • templates • playbooks • case studies • experiments And every post had the same CTA: “Comment X and I’ll send it.” That single mechanic generated thousands of inbound leads. Every week, each team member created: • 1 new lead magnet • multiple distribution angles from it We also added: • free tools • customer stories • motion design videos • better landing pages New hire: 1 Product Manager. STEP 4: Layer distribution channels At this point, the system already worked. Now the goal became: add more attention sources. We expanded into: • X/Twitter • LinkedIn influencers • newsletter sponsorships • partnerships But we never stopped the channels that already worked: • outbound • Reddit • LinkedIn content Most founders abandon winning channels too early. We scaled them harder instead. STEP 5: Scale paid acquisition Once the organic engine was stable, we added: • Meta Ads • Google Ads • UGC creators • B2B influencer campaigns Then we redesigned the entire website around conversion. At the same time: we scaled hiring across: • growth • engineering • sales And massively increased outbound volume with GojiberryAI. Our philosophy is simple: More targeted attention → more conversations → more demos → more customers Most startups die from lack of distribution, not lack of product. Build distribution earlier than everyone else. That changed everything for us. If you want to try GojiberryAI with a 14-day free trial instead of 7 days: Comment “GOJI” and I’ll send you access.