AI Should Not Have a Meter: The Case for Local Inference
Guilherme Moreira
929 views
Spencer Thomason (@STARTUPHAKK_sh) has been writing .NET since the first beta, built for GoDaddy and JDA, and founded around ten companies of his own. Today StartupHakk runs on three engines: a YouTube channel doing five million views a year, a software
consulting firm, and a state-licensed apprenticeship that trains developers and deploys them into real client work.
Two months ago he launched Open Mono Agent — an open-source coding agent that runs entirely on your own hardware. No subscription, no data leaving your network, no token billing. His framing: AI shouldn't have a meter.
We get into why he thinks vendor lock-in to a frontier model is a liability rather than an advantage, how a 35-billion-parameter model with a good harness closes most of the gap on a two-trillion-parameter one, and why the real opportunity right now isn't building AI — it's connecting the systems companies already have. His phrase for it: you need IA before you need AI. Infrastructure
architecture before artificial intelligence.
Also: why he killed the bootcamp model, why 8,000 pieces of content is what five million views actually costs, and why he thinks high interest rates have hurt junior developers more than AI has.
This episode is sponsored by Inevitable Flow. Inevitable Flow helps B2B founders and executive teams in high-trust industries fix their pipeline, automate process, and build predictable revenue engines — so growth doesn't depend on referrals, luck, or the founder doing everything.
Start a conversation: https://calendly.com/gui-moreira/30min
CHAPTERS
0:00 Intro
1:44 25 years on .NET, and what made him go independent
2:22 Building Clean Router on nights and lunch breaks
4:42 What ten companies taught him about failure
5:02 How to test product-market fit for $500
6:58 The three engines behind StartupHakk
9:49 8,000 videos to reach five million views a year
10:45 Why consulting scales badly
12:44 Killing the bootcamp model
14:16 The rise of the forward-deployed engineer
18:31 "You need IA before you need AI"
19:20 Open Mono Agent and the case against a meter
20:10 What runs on $600 of hardware
25:16 What you actually give up going local
28:00 Why the harness matters more than the model size
30:20 Vertical models, not AGI
33:08 How an open-source project pays for itself
39:04 Is AI a threat to developers?
43:46 The Magic Wand question
SPENCER THOMASON
StartupHakk: https://startuphakk.com/
Open Mono Agent: https://github.com/StartupHakk/OpenMo...
YouTube: @startuphakk_sh
GUI MOREIRA
LinkedIn: LinkedIn: guilhermesvsmoreira
Inevitable Flow: www.inevitableflow.com
#AI #SoftwareDevelopment #OpenSource #LocalLLM #TheInevitableCast
AI Should Not Have a Meter: The Case for Local Inference
Guilherme Moreira
929 views
Spencer Thomason (@STARTUPHAKK_sh) has been writing .NET since the first beta, built for GoDaddy and JDA, and founded around ten companies of his own. Today StartupHakk runs on three engines: a YouTube channel doing five million views a year, a software
consulting firm, and a state-licensed apprenticeship that trains developers and deploys them into real client work.
Two months ago he launched Open Mono Agent — an open-source coding agent that runs entirely on your own hardware. No subscription, no data leaving your network, no token billing. His framing: AI shouldn't have a meter.
We get into why he thinks vendor lock-in to a frontier model is a liability rather than an advantage, how a 35-billion-parameter model with a good harness closes most of the gap on a two-trillion-parameter one, and why the real opportunity right now isn't building AI — it's connecting the systems companies already have. His phrase for it: you need IA before you need AI. Infrastructure
architecture before artificial intelligence.
Also: why he killed the bootcamp model, why 8,000 pieces of content is what five million views actually costs, and why he thinks high interest rates have hurt junior developers more than AI has.
This episode is sponsored by Inevitable Flow. Inevitable Flow helps B2B founders and executive teams in high-trust industries fix their pipeline, automate process, and build predictable revenue engines — so growth doesn't depend on referrals, luck, or the founder doing everything.
Start a conversation: https://calendly.com/gui-moreira/30min
CHAPTERS
0:00 Intro
1:44 25 years on .NET, and what made him go independent
2:22 Building Clean Router on nights and lunch breaks
4:42 What ten companies taught him about failure
5:02 How to test product-market fit for $500
6:58 The three engines behind StartupHakk
9:49 8,000 videos to reach five million views a year
10:45 Why consulting scales badly
12:44 Killing the bootcamp model
14:16 The rise of the forward-deployed engineer
18:31 "You need IA before you need AI"
19:20 Open Mono Agent and the case against a meter
20:10 What runs on $600 of hardware
25:16 What you actually give up going local
28:00 Why the harness matters more than the model size
30:20 Vertical models, not AGI
33:08 How an open-source project pays for itself
39:04 Is AI a threat to developers?
43:46 The Magic Wand question
SPENCER THOMASON
StartupHakk: https://startuphakk.com/
Open Mono Agent: https://github.com/StartupHakk/OpenMo...
YouTube: @startuphakk_sh
GUI MOREIRA
LinkedIn: LinkedIn: guilhermesvsmoreira
Inevitable Flow: www.inevitableflow.com
#AI #SoftwareDevelopment #OpenSource #LocalLLM #TheInevitableCast