If you only read one thingHow to keep building your engineering skills as a CTO (9 minute read) Should a CTO be able to code? Should she code? Age-old question. Many good points in the article to stay more technical (also for the joy!). I often recommend to my coachees to âBuild proofs of concept on technologies I want to know more aboutâ - this brings joy and has impact, as long as you have a defined process to merge this with the product feature stream. Donât delegate everything that brings joy as a CTO. https://kaiomagalhaes.com/blog/Staying-technical-as-a-CTO ComicFrom CommitStrip https://www.commitstrip.com/en/2012/08/14/pair-programming/ Also see pair-programming below. Tweet of the weekfrom https://twitter.com/karinanguyen_/status/1764789887071580657 Stories Iâve enjoyed this weekLeadership Is A Hell Of A Drug (21 minute read) Be warned strong language in the article Funny read. No really. âYou have been invited to: Compulsory On-Siteâ Donât be that manager, Especially talking to developers with marketing corporate speak instantly annoys them. Be real. âFirstly, why did you clowns make it a compulsory on-site event? Do you really think that it was worth dragging everyone into the office against their will because we needed to listen to you talk for four hours in a row?â And the one thing I wonder all the time myself âwhy did you need someone from outside the organization who doesnât know a damn thing about our work to come in and tell you how youâre screwing up?â Like those management consultants telling the CEO of a steel factory how to produce steel. https://ludic.mataroa.blog/blog/leadership-is-a-hell-of-a-drug/ Behind the feature: The multiple lives of multi-edit (17 minute read) Many of my CTO clients complain that features from product are not deep or challenging enough. This is a deep feature, and if you need something to point to, point to this. Many product teams I know would have stopped at the first UI mockup. https://www.figma.com/de/blog/behind-the-feature-the-multiple-lives-of-multi-edit/ Humans are not automatically strategic (10 minute read) âPerhaps 5% of the population has enough abstract reasoning skill to verbally understand that the above heuristics would be useful once these heuristics are pointed out.â And a lot of these 5% are in development. Other people are different. Understanding this is important for every techie. Boing, very important article. Plus: Too many CTOs are not strategic enough, e.g. they donât have any tech strategy at all. âI know people who are far more strategic than I am, and there seem to be clear avenues for becoming far more strategic than they are.â I want to become more strategic too, so should you. Take those avenues. https://www.lesswrong.com/posts/PBRWb2Em5SNeWYwwB/humans-are-not-automatically-strategic A list of small teams that ma(d|k)e a lot of money or were acquired for a lot of money. Instagram, 13 employees, $1billion exit. We have not yet harnessed the power of small teams. What could you do with a small team if it wasnât for all the boring features someone cooked out that have no impact? https://stevepulec.com/posts/small/ A new study finds that the later we meet someone in a sequence, the more negatively we describe them (5 minute read) OMG. So itâs good to be first I guess. Which fits me (one of my faults is Iâm too competitive). Me aside, itâs good to keep this in mind when hiring! Or perhaps when you have a string of 1on1s. Perhaps. https://suchscience.org/people-encountered-later-in-a-sequence-described-more-negatively/ An Overview of the LoRA Family (20 minute read) LoRA is to tailor LLMs, âLow-Rank Adaptation (LoRA) can be considered a major breakthrough towards the ability to train large language models for specific tasks efficiently.â For everyone adapting their own LLM, this is essential. LoRA, LoRA+, VeRA, LoRA-FA, do you know what they mean? If you donât want someone else to be Chief AI Officer, this is for you. https://towardsdatascience.com/an-overview-of-the-lora-family-515d81134725 What if we rotate pairs every day? (22 minute read) Pair programming is undervalued. CTOs ask me what to do to get higher cohesion when everyone is remote? Well, start with pair programming. And as the article implies, rotate pairs every day (not so sure about that myself, but they did an experiment, I didnât). Do you do pair programming in your department? Do you mandate it? https://martinfowler.com/articles/rotate-pairs-experiment.html Training great LLMs entirely from ground zero in the wilderness as a startup (13 minute read) Everyone says âTrain your own LLMâ, do AI the right way (I tell this everyone). But how do you do this? Excellent article on the topic to get you started and also to make a leap forward. https://www.yitay.net/blog/training-great-llms-entirely-from-ground-zero-in-the-wilderness The Dumber Side of Smart People (7 minute read) This article is about you and me: âBeing very smart makes it harder to listen to people who are less credentialed than you, even when they might have the right answer.â I might remove the âcredentialedâ, but even that might be ok, if we consider all developers having credentials and everyone outside tech not so. I find it very difficult listening to people when I already know where this leads (same with a movie when itâs clear after 15 minutes how it will end). But Iâve learned to calmly listen to people because Iâm often wrong, which I could have found out by listening better. https://collabfund.com/blog/the-dumber-side-of-smart-people/ Employee Retention Tips for HR Managers (10 minute read) Ah, retention. Probably not a topic for a time when everyone fires developers - ahem, lays off, but I guess it doesnât make a difference to those shown the door - Yes, itâs trivial at its core, Watson, âit is about fostering an environment that encourages employees to stay, contributing to a stable and productive workplaceâ instead of an environment, that encourages people to leave. So simple, but to many companies this simplicity is lost. Besides this, some more insights in the article. And always remember, before filling water into a bucket, close the holes. https://mybites.co/blog/hr/employee-retention-tips-for-hr-managers Using LLMs to Generate Fuzz Generators (14 minute read) Developers donât use fuzzers enough. And itâs not only for C or Rust. But for every programming language to find problems in your code (even webapps can benefit from fuzzing). And LLMs will supercharge fuzzing and bug finding. Although the article is sceptical, Iâm seeing the positive bits of the article âFor each vulnerability the LLM identifies, ask it to generate a directed fuzzer that generates inputs to try to trigger (just) that vulnerability.â https://verse.systems/blog/post/2024-03-09-using-llms-to-generate-fuzz-generators/ Signs youâre about to be acquired (7 minute read) Youâre part of the leadership team, so you should know. But because you think youâre handling this acquisition like a stealthy ninja, developers will suspect there is something going on. Join the CTO newsletter! | |