# VC thoughts, Palchinski & (not so) fun errors

# Where’re software (mostly) & deep tech VC headed \[2\]?

![Curiouser and Curiouser | Family Wise](https://family-wise.co.uk/wp-content/uploads/2016/03/V309-Curiouser-and-Curiouser-flat.jpg align="left")

Just some additional **thoughts** in addition to a [recent post](https://posts.teleogenic.com/metha-venture-in-2025):

* The cost of developing an MVP/PoC is declining rapidly ([vibe coding](https://andrewchen.substack.com/p/predictionsthoughts-on-vibe-coding) much?); besides the hype and despite some IMO [poisonous practices](https://www.indiehackers.com/post/starting-up/vibe-coding-for-dummies-11-tips-for-building-with-ai-HlxaY8YDHlI3SaDiEeY0?utm_campaign=ih-newsletter&utm_medium=email&utm_source=indie-hackers-emails) (how does *‘Start over, often’*, *‘Don’t think about architecture’* sound?) advised, it’s [changing the landscape with baby steps](https://pocket.co/share/3cd022bf-fece-4c7a-824c-d6c34c4deb06)
    
    * in addition to the last link author’s thoughts, I’d add that some **visual-code primitive frameworks** are likely to emerge → i.e. an adaptive page/app that is generated in markdown-like blocks that are rendered on-the-fly. The bottleneck here is essentially tokens/s output, and that paradigm would supplant it.
        
    * Yeah, definitely **testing and security/hacks**. It could be a vibey TDD, e.g., a *‘define outcome → get tests → code is built around them’.*
        
* By Jevon’s paradox (and as majority will adopt the ‘Start, prompt, re-do’ approach), we’d need:
    
    * more **energy** to support this (e.g. the [bid on nuclear](https://www.theinformation.com/articles/can-tech-get-nuclear-power-to-move-at-ai-speeds?utm_source=pocket_saves); but let’s not count hydrogen/alt energy out!), along with
        
    * more efficient **means of computation** (mentioned different [emerging paradigms in a recent post](https://posts.teleogenic.com/ai-in-scientific-discovery))
        
    * efficient **algorithms and architectures**, and some safe haven of **good code**
        
    * efficient **infrastructure** (data transfer, storage, etc.), so - the usual ‘boring’ areas: apart from something that works, it has to be computed, stored, produced
        
* VCs are definitely going to reflect that - one most likely wouldn’t need 200K-1M to build a PoC soon. So,
    
    * ![Venture Capital Investment Market Report 2025, Share and Size By 2034](https://www.thebusinessresearchcompany.com/graphimages/Venture_Capital_Investment_Market_2025_Graph.webp align="center")
        
        less funded **software products/micro-products** (i.e. [one doesn’t need inflated valuations](https://avc.xyz/drinking-from-a-firehose))
        
    * more **deep tech**/complex products → bigger average rounds
        
    * **cycles getting shorter for even traditionally ‘slow’ fields** like healthcare and material science (pushing \**a-lot* more iterations in the same time + faster testing/simulation/adoption processes)
        
    * outreach and events/networking becoming **more important than funding** for software
        
    * **VCs specialising** in e.g. scouting, GTM and scaling, network effects and collabs, advocacy/outreach/media + founders looking for a ‘portfolio’ of funds
        
    * **smaller teams** overall?
        
* Deep tech needs expertise + as I’ve mentioned before → [more VC funds](https://avc.xyz/drinking-from-a-firehose) point to **dwindling returns and the need to spot projects/teams early on**.
    
    * So, even more hackathons/events to nurture and push the little starlets?
        
    * Prediction/consensus/fact-checking markets?
        

## Bottom line

It’s all going to become **more and more interesting, faster.**

# [Palchinskii’s principles](https://powermaps.net/tpost/fyjza7kre1-chapter-3-palchinskiis-principles)

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1742130830864/f5f3ad36-79f8-447b-9f2c-eaa64e00e670.png align="center")

(Even adopted at the [US military](https://juniorofficer.army.mil/the-power-of-the-palchinsky-principles-how-failure-can-lead-to-success/)! Wooo)

I can recount Marcus Guest giving a lecture at my university a long time ago, which was telling ‘the experiments should be [safe to fail](https://freedium.cfd/https%3A%2F%2Fmarcusguest.medium.com%2F28-success-requires-moving-from-fail-safe-to-safe-fail-66858d23ca0f), and one should [cross the river by feeling the stones](https://en.wikipedia.org/wiki/Crossing_the_river_by_touching_the_stones)’. So, **Palchinski’s principles** lay the foundation with iterative experimentation, maximizing risk:reward ratio, [kaizen](https://en.wikipedia.org/wiki/Kaizen)\-style improvement. It also

## **Experiment with a *variety* of ideas**

to increase the chances of success.

## **Accept that some failure is inevitable,**

so keep projects small enough that failure is *survivable.*

## **Develop effective feedback loops**

between decision-makers and those closest to the action to quickly identify and *select* what works in the local context.

# Fun news

## [Minority report is scary? How about 17 months in jail over an AI mistake…](https://www.abc27.com/national/man-jailed-over-police-ai-program-then-freed-17-months-after-victim-raised-doubts/amp/)

![Jailed over police AI program, then freed 17 months after victim raised  doubts](https://i.ytimg.com/vi/Aw8YxFdEcG8/maxresdefault.jpg align="center")

So, Chris Gatlin didn’t have a lucky day, or week, for sure; he was misidentified by a facial recognition algorithm and was in jail for 17 months before the court realized there was no hard evidence.

## [A slightly better news - someone’s won 81T in a bank error lottery 🤗](https://sherwood.news/business/citigroup-accidentally-gave-a-client-usd81-trillion-instead-of-usd280)

![Chris Reynolds' PayPal account was erroneously credited  $92,233,720,368,547,800.](https://media.cnn.com/api/v1/images/stellar/prod/130717021713-paypal-error.jpg?q=h_501,w_584,x_0,y_0,c_crop/h_576,w_1024,c_lpad,b_rgb:061015/h_833,w_1480 align="center")

Instead of $280, a client was ‘sent’ $81T, which is 500+ times more than Citibank’s market cap ❣️

The fun thing is that the error was caught by a **third employee** checking, **90 minutes after** the initial error. Raises some questions about the financial system and the banks’ risk management practices <3

All I can remember is PayPal [erroneously sending a client $92 quadrillion in 2013](https://edition.cnn.com/2013/07/17/tech/paypal-error/index.html)…