ACT (Almost Cringe Tech)

Miguel Twahirwa and Franck Nijimbere

0.0 (0)
Technology
Semiweekly

Two builders on what’s next in technology and how to actually take advantage of it. Miguel Twahirwa and Franck Nijimbere talk AI, shipping products, finding customers, and the uncomfortable work of getting better—with honest disagreements and almost-cringe experiments along the way. New episodes Tuesdays and Thursdays.

1d ago

I Lost the Hackathon on the Demo

Miguel spent 24 hours at Boson AI's Higgs Audio Hackathon building Genie & Rookie, two learning companions that help a child explore an idea instead of just hearing an answer. He walks Franck through the demo, then admits what cost him: he forgot the cardinal rule of hackathons, that demos matter more than what you build. They break down the projects that won, how AI is changing interviews, and what hiring managers told Miguel: you see more AI usage among really good engineers and really bad engineers. Then the taste debate. Franck argues taste is the moat, because a lot of people aren't applying much taste to what they ship, and explains how GoFaceless routes requests it can't handle yet to a worker that figures out the feature. Miguel shares his design process, from building a library of things he finds beautiful to iterating on images before he writes code. They close on Franck's plan to have an AI clone of himself apply to YC. "To be able to produce something, you have to be able to eat." (Miguel) Chapters 0:00 Cold open 0:30 24 hours at the hackathon 2:37 Miguel's demo: Genie & Rookie 4:54 Demos matter more than what you build 5:38 What won the hackathon 14:43 Interviews in the AI era 19:34 Good engineers, bad engineers and AI 21:58 Is taste the only moat? 26:41 Turning requests into features 30:26 Miguel's design process 37:39 Franck's AI clone 41:20 Wrapping up ACT (Almost Cringe Tech) — Episode 9 Hosts: Miguel Twahirwa and Franck Nijimbere.
3d ago

It’s Cringe Not to Be Cringe

GoFaceless's ChatGPT app was accepted, and it's already sending Franck 20 to 40 visitors a day. Churn is down to about 45%, from about 50%. Franck explains how ChatGPT apps work (you can't sell inside them, so you send people back to your platform) and why building them now feels like being early to the App Store. Then ads: on his best-performing ads, a trial costs him $2 to $4. Miguel makes the case that you have to be willing to look bad before you get good, and Franck goes further: "it's cringe not to be cringe." That turns into the disagreement at the heart of the episode. Franck keeps GoFaceless looking as established as possible and won't put his own story front and centre, because he'd be fighting too many biases. Miguel argues that founder-led content and authenticity win anyway. Franck's answer: "Systemic bias is definitely an invisible glass, but it's real." "It's cringe not to be cringe." (Franck) Chapters 0:00 Cold open 0:23 GoFaceless by the numbers 4:39 How ChatGPT apps work 7:32 Early to a new app store 11:45 Which ads work 15:19 Growing on TikTok 19:39 You have to look bad first 21:41 It's cringe not to be cringe 24:01 Do we have to do the dances? 28:25 Does where you're from hurt the brand? 40:26 Systemic bias is real 44:26 Wrapping up ACT (Almost Cringe Tech) — Episode 8 Hosts: Miguel Twahirwa and Franck Nijimbere.
Nov 5

Should Everyone Get an Uncensored AI?

Railway, the hosting platform behind GoFaceless, was breached, and Franck spent a day rotating close to fifty keys after Stripe flagged one of them. Miguel spots the opportunity: most PHP sites belong to small businesses that don't know they're exposed. Franck had already shipped a PHP patch site, which got 24 visitors in its first 24 hours before Google had even indexed it. From there: why Miguel only builds what has a marketing path ("think marketing first"), counter-AI and the return of real-life community, and the big debate. An "obliteration lab" lets anyone strip the guardrails off open models. Miguel worries about who gets the cyber-weapons; Franck argues that it's either no one or everybody, and that agents need identities (KYC for agents). They close on writing like a human: Breaking Bad's hooks, frameworks, and organized chaos. "A good framework is invisible." (Miguel) Chapters 0:00 Cold open 0:23 Railway got breached 8:54 Franck's PHP patch site 13:10 Passion, or a path to market? 16:44 Marketing first, product second 18:38 Counter-AI and real-life community 22:32 Should everyone get an uncensored AI? 30:01 KYC for AI agents 33:17 Writing like a human 41:07 Wrapping up ACT (Almost Cringe Tech) — Episode 7 Hosts: Miguel Twahirwa and Franck Nijimbere.
Nov 2

The Skills That Matter After AI Improves Itself

If AI starts improving itself, which human skills still matter? Franck says agency, being prolific and being visible. Miguel picks storytelling. They get into ingestive versus generative people, what Pixar knows about stories, and why some people are funny. The episode closes with a live working session on GoFaceless: 30 to 50% trial-to-paid before ads, 50 to 60% churn, and a pricing test on PostHog. "If I were to pick one skill, I would say storytelling." (Miguel) Chapters 0:00 Cold open 0:18 The skill that matters after RSI 6:05 Ingestive or generative 7:51 Learning to tell stories 12:16 What Pixar knows 16:52 Why some people are funny 21:53 Something slightly cringe 25:08 Fixing 60% churn
Nov 2

How to Actually Learn AI in 2026

Why learn AI from scratch when models are becoming a commodity? Miguel makes the case for building an LLM yourself, they ask whether any of it matters after recursive self-improvement, and they talk about privacy, models "losing their humanity," Claude and the Church, and why Africa needs its own lab. "It's a massive responsibility." (Franck, on building our own models) Chapters 0:00 Cold open 0:29 Down the MFM rabbit hole 6:29 How we learn 12:23 Building an LLM from scratch 14:34 Does it matter after RSI? 20:35 Privacy and functional masking 24:11 Models losing their humanity 29:41 Claude and the Church 33:16 Why Africa needs its own lab
Nov 2

Will Today's Tech Giants Still Win in 10 Years?

The clearest disagreement on the show so far. Franck: "I don't think that new winners will be created." Miguel: "I'm not so confident that Anthropic or OpenAI is de facto the king of AI." Along the way they cover an agent that fought an airline, who actually uses AI, OpenAI's math drop, whether labs will hide breakthroughs, and whether technical moats still exist. Franck also shares his output on days with full electricity: around 300 commits in a day. "Agency times resources." (Franck) Chapters 0:00 Cold open 0:26 Inbox zero and AI assistants 5:33 An agent fights an airline 11:58 Who actually uses AI 15:51 OpenAI's math drop 21:51 Will labs hide breakthroughs? 31:26 Is there still a moat? 39:26 Will today's winners keep winning? 53:52 Threads or X
Nov 2

China in 5 Years. Africa in 50?

AI moves fast, but people adopt it slowly. Franck says "code does not matter anymore," Miguel says there will always be room for b******t artists, and they debate whether consultants survive AI. Then they put timelines on it: AI-run hotels in China within five years for Franck, more like 30 for Miguel, and closer to 50 for Africa. "Most of the time technology is not the issue. Humans are typically what gets in the way." (Miguel) Chapters 0:00 Cold open 0:32 Trying new things 3:29 Work with nothing to show 11:00 Adobe, open-sourced 18:13 Who software is for now 26:11 Do consultants survive AI? 32:04 How much has really changed? 41:34 Physical AI and self-driving hotels 55:46 Virtual economies
Nov 2

90% of Our Users Click "Surprise Me"

GoFaceless now has nine or ten templates, and none of them has beaten Hotel Lobby yet. Franck shares a surprising number: about 90% of users press the "Surprise me" button instead of bringing their own idea. That leads into why people share, what happens when the familiar meets the unexpected, growth loops (including Tuckwell's gift loop), how the two of them pick what to build, and B2B versus B2C. "B2B was actually the best place to be." (Franck) Chapters 0:00 Cold open 0:14 Building templates for GoFaceless 7:23 Familiar plus unexpected 12:41 Why people share 16:40 Building growth loops 21:03 Tuckwell and LineUp 29:41 How we pick what to build 44:12 B2B or B2C 50:59 Support as a growth channel
Nov 2

Going Viral in Burundi, and Why People Hate Good AI Art

Franck's GoFaceless video of two well-known Burundian actors went viral: about 75K views on Instagram and roughly 1,000 to 2,000 new users in five days. From there, Miguel and Franck get into the backlash against AI art. Miguel argues that well-made AI art is still art, because "the whole point of it is to make people feel." Franck pushes back that "too perfect is not human." They also cover the watermark trade-off, a launch video Miguel made for about $2, and whether AI is eating taste. "Being intentional is going to be a moat." (Miguel) Chapters 0:00 Cold open 0:23 Going viral in Burundi 8:25 The watermark trade-off 12:36 The AI art backlash 20:35 Intention is the moat 25:38 A launch video for $2 36:42 Is AI eating taste? 42:34 Can models escape the average? 49:04 How to build taste
Nov 2

Almost Cringe Tech — Start Here

Two builders. Honest disagreements. Almost-cringe experiments. Meet ACT (Almost Cringe Tech), with Miguel Twahirwa and Franck Nijimbere: conversations about AI, shipping products, finding customers, and the uncomfortable work of getting better. This trailer features moments from episodes 7–10. Full episodes arrive Tuesdays and Thursdays at 10 a.m. Toronto time, starting November 5. Follow ACT to get the next conversation. Upcoming: uncensored AI, founder visibility, the hackathon demo, and what happens when software approaches zero. Watch ACT on YouTube: https://www.youtube.com/channel/UCf7sXJSv3_wmVYy0868FbAQ

1 Trailer

9 Episodes

Trailer

Almost Cringe Tech — Start Here

Two builders. Honest disagreements. Almost-cringe experiments. Meet ACT (Almost Cringe Tech), with Miguel Twahirwa and Franck Nijimbere: conversations about AI, shipping products, finding customers, and the uncomfortable work of getting better. This trailer features moments from episodes 7–10. Full episodes arrive Tuesdays and Thursdays at 10 a.m. Toronto time, starting November 5. Follow ACT to get the next conversation. Upcoming: uncensored AI, founder visibility, the hackathon demo, and what happens when software approaches zero. Watch ACT on YouTube: https://www.youtube.com/channel/UCf7sXJSv3_wmVYy0868FbAQ

Two builders on what’s next in technology and how to actually take advantage of it. Miguel Twahirwa and Franck Nijimbere talk AI, shipping products, finding customers, and the uncomfortable work of getting better—with honest disagreements and almost-cringe experiments along the way. New episodes Tuesdays and Thursdays.

Creator

Miguel Twahirwa and Franck Nijimbere
Years Active

2026
Episodes

10
Rating

Clean
Copyright

© Miguel Twahirwa and Franck Nijimbere
Show Website

ACT (Almost Cringe Tech)