Consumer underwriting time decreased by 94% since March
Michael Lee describes building a system that can take on all consumer loans within the organization, decreasing average consumer underwriting time by 94% since March.
40,153 results · Most relevant
Michael Lee describes building a system that can take on all consumer loans within the organization, decreasing average consumer underwriting time by 94% since March.
Speculative decoding can save a lot in inference generation cost, but it is not free: hosting two models requires accounting for the memory needed to host the second (draft) model, plus extra KV cache space for both models.
Nikhil explains that dynamic GPU reallocation between training and inference is only possible when everything runs in a single infrastructure stack. When user load spikes and inference needs to burst, hardware can be reallocated from the training cluster to the inference service; when a Slurm job finishes, inference pods can be scheduled on those GPU nodes immediately through Kubernetes because they are still just nodes in a pool.
deterministic
Eugene describes his ecosystem: short form content is set up to gather leads, with actual conversions (sales) preferred outside of short form — on long form where viewers have more time to build trust.
The speaker defines the ideal guest profile as websites that are not super authoritative but have significant search traffic, sets the campaign to target 200 prospects, and sets contact title priorities in the outreach tool settings so it looks for people with those titles as the first contact point. He starts the ICP interview by explaining why he's doing outreach — his video series Traffic Hunters, where he interviews founders about marketing secrets.
The speaker warns that AI-generated agent instruction files need careful review. When he went to review the files, something was clearly off: the first pass contained errors requiring multiple rounds of fixes, including a corrupted file that had to be regenerated. His approach was to ask for a bullet point list of all required fixes and iterate until the files were in a good place.
The speaker set a specific success criterion before running the outreach campaign: booking at least five meetings. To test the pipeline end-to-end, he prompted Slack to kick off the campaign and start asking questions from the ICP interview, watched for the campaign coordinator to spawn the prospector after the brief was filled out, and ran the copywriter one last time as the final test after fixing the vetter.
The speaker explains his tool selection process. He ruled out N8N quickly because it feels like Zapier — dragging and dropping steps into a visual workflow rather than AI-native building. He chose OpenClaw because it is not tied to one model (allowing him to plug in tools like Nano Banana for image generation), and because it can run through Slack, letting him work inside the tool he already uses all day. He narrowed his options to three or fewer before committing.
The speaker spent hours debugging the vetter, tracing the specific data field causing the error — the traffic slope field showing negative values for sites with actual growth. He chose to fix the core issue in the tool rather than working around it, spawned sub-agents to work on all the fixes while he debugged in parallel, and finally things were better. He verified the fix by refreshing the dashboard and confirming the vetter now correctly identifies growing traffic trends.
Hormozi argues that many people in the fitness and aesthetics world believe that becoming a pro or winning shows will make them successful in business, which is far from the truth. There are tons of broke fitness people; some past champions who didn't learn business are broke — they get their moment in the sun, their sponsors go away, and then they're back to zero again.
Hormozi observes that fitness competitors stick to what they know how to improve (reps, eating, cardio) and choose not to be beginners in other areas like business, marketing, and sales. They unconsciously dedicate five to ten-plus years of their life to their sport and take years off their life health-wise, without getting what they want — because it was emotionally uncomfortable to think about sucking at something new.
The speaker automated the parts of outreach that were eating his time, and says his AI outreach agent gets better response rates than an expensive agency they hired in the past. He adds that this outreach work is what companies pay agencies for month after month, and the AI system beats agency response rates because it gets to know the objectives and business firsthand and executes better than most pros.
stated
The speaker frames the five personality colors as a tool: identify which color you are talking to before choosing your conversational approach. He illustrates this by noting that being too gamey with a shy girl would actually work with a social butterfly, so tactics must match the personality color.
The host argues that most men wait for very big signs of interest before moving forward, and that this is a problem precisely because women don't show interest the same way men do — so waiting for big green lights means missing the subtle signals that are actually there.
Anish Acharya argues distribution has always been a moat but is now an increasingly massive advantage, especially the ability to get your product into people's feeds and keep them reminded of it.
Anish Acharya argues the failure of AI consumer products is a product design failure, not a capability failure — the opportunity is a product design challenge, not a model or capability challenge.
The creator argues the economics are changing: the cost of creating code will drop, but the cost of creating good systems will increase, and the cost of appearing in front of buyers at the right moment will also increase. These are factors founders need to think about.
The guest claims retail stores have drastically lowered their customer acquisition cost — a fraction of what it costs online.
Matt Paulson explains that when you can't afford to participate in an exclusive opportunity, you can reframe it: the people who can't afford to be in the room probably care more about the experience than those who can. Making the opportunity available online to a much wider audience — recognizing that a hero like Magic Johnson has fans around the world, not just in that room — can raise far more money and awareness, and becomes a business model built on that reach.
None stated.
Around two and a half minutes into the interaction, start thinking about what she's up to right now, how much time she has, and whether you can lead her somewhere (like to get a coffee or tea).
Choose sources → to get a library of your own and a daily brief.