Weekly digest #86

We (Matt and Shray) read hundreds of articles on company building, venture investing, and self-management and curate the best ones into a weekly digest to help founders and operators stay on the top of their game.

Agent psychosis (10 minute read)

AI coding agents create a dopamine loop that makes people feel hyper-productive while generating massive amounts of low-quality code. People develop parasocial relationships with their AI, prompt it without critical thinking, and flood open source projects with submissions that take minutes to generate but hours to review. The asymmetry is brutal for maintainers. Some projects now prefer receiving prompts over actual code because there’s more trust in running the agent yourself than accepting someone else’s AI-generated output.

How metrics change your values and behaviour (38 minute read)

This guy produces gems. Metrics motivate. Steps, likes, grades, scores, etc. They record a value, yes, but they encourage a behaviour. If you’re not careful, you can get swept up chasing metrics (and exhibiting behaviours) that don’t really serve you. People optimise for engagement on Twitter instead of meaningful communication, chase Fitbit step counts over actual health, or farm law school rankings instead of learning. Audit the metrics you care about ruthlessly.

The Palantirization of everything (13 minute read)

Palantir’s model of embedding engineers inside customer orgs to build custom workflows looks attractive, especially as enterprises struggle with AI adoption. But most startups copying this will become expensive services businesses with software valuations. Palantir works because there’s a real platform underneath the bespoke work, built on reusable primitives. They also operate in mission-critical domains with huge contracts that justify the cost. Most startups lack both the platform spine and the high-stakes problems needed to make this model work at scale.

Dinners are the new trade shows. Here’s how to run them well (15 minute read)

Ah, the exec dinner. Wine and dining sounds Machiavellian, but in the context of B2B sales, it really works. Often much more profitable than a large trade show. Instead of paying for a booth and hoping the right person walks by, you can handpick 12 or so people in your pipeline or target list. But there’s a strategy to doing them well. Some core principles anchor the whole thing, though. Be intentional about your guest list (the guests should want to meet each other). Centre the dinner around a theme of reason. The vibe of the venue has to match your brand.

My favorite GTM tools and ones that i want to try in 2026 (8 minute read)

There’s many tools out there now for marketing, sales, and revops. Here’s reasonably good list of GTM tools to help you with ABM, retargeting, inbound, emails, and demos. Phantombuster and Clay make powerful allies. Happy to share my own workflow (Matt) if you’re curious, just hit reply. And if anyone has figured out AEO (in a none hack way) - let me know.

Refer a friend

Refer and we’ll send you our favorite books as a “thank you” for spreading the word.

Who will win, legacy SaaS or AI-native companies? (2 minute read)

Legacy SaaS like Salesforce has all your company context but can’t act autonomously. AI startups like Decagon can automate tasks but lack context about your business. The massive opportunity is combining both so you can automate entire workflows. The race is whether legacy SaaS builds agents first or AI startups accumulate context first.

The lost signal (3 minute read)

When capital was scarce it meant a lot to raise it. VCs would do a lot of DD which, in a way, would validate the opportunity. But capital isn’t scarce anymore. VCs deploy way faster, often with much less DD. You can’t rely on them to help validate your idea anymore. You’ve got to do more of the background/early work yourself. Which sounds obvious; “why would I dedicate my time to something I don’t believe in?”, you ask. Having belief and true conviction are distinct things. The latter requires answering questions like how many novel assumptions does this rely on, what will it cost to prove them, and what’s the ultimate payoff if it works, etc.

Silicon Valley’s 99% blindspot (6 minute read)

Rex argues that Silicon Valley builds products for itself (the 1%) and misses huge opportunities serving everyone else. Tech founders gravitate toward problems they personally experience. There’s a much bigger opportunity outside of that circle. I partially agree - but you still see lots of startups in the consumer tech space benefitting the everyday person. But perhaps there’s more we could be doing in B2B for the 99%.

Move faster (5 minute read)

You’re probably too slow. While building software, we’re conditioned to expect friction, to wait for builds, to schedule meetings, to ‘sleep on it’. In the current era of AI, when intelligence is on tap and initiation is near-instant, waiting starts to become a choice, not a necessity. In a slow world, we use the ‘waiting time’ of the process to subsidize our own lack of clarity. In a fast world where the cost of execution collapses to near-zero and the fundamental physics of how you build things changes, that subsidy is gone. If you aren’t moving, it’s not because the system is compiling; it’s because you don’t know what to do next. By moving fast, you increase your ‘luck surface area’. You take more shots on goal, test more hypotheses, and stumble into more happy accidents.

No XP waste (2 minute read)

There’s a phrase only keen lobster fisherman will know; “no XP waste”. Runescape was a classic from the early 2000s. It was such a grind of a game, though. Endlessly clicking your mouse to chop wood, fight orcs, or yes, catch lobsters. Players of the game would use appscripts to have their character do these tasks while they were AFK. Hence; “no XP waste”. In other words, you’re always leveling up your character around the clock, no matter what you’re doing. The same phrase is being applied to orchestrating agents in knowledge work. Constantly have them doing background tasks - if you don’t, well, you’re wasting XP.

Why you should prepare for a layoff (5 minute read)

Even if you’re not anticipating layoffs, it’s good to be prepared. Here’s a checklist from Steve, a long time engineering leader at Amazon. It’s nothing you’d be surprised to learn, but a nice reminder. Have a savings buffer, update your CV while you’re at your job (so the info is fresh), always take note of your impact (the metrics and projects), and keep your network warm (but don’t be needy).