Screenshot from an interaction with a LLM where the answer was “You’re right”, instead of the infamous “You are absolutely right”.
It turns out that I am simply "right". Why I am not "absolutely right” like everyone elseSc??? This is outrageous. 😠
«There’s an irresistible, almost demoralizing irony in the fact that developers are discovering docs and accessibility only now due to AI»
passo.uno/skills-are-d...
«I hope nobody ever uses that thing or would try to build an actual browser out of it, at least with this generation of agents, it’s still pure slop with little oversight. It’s an impressive research and tech demo, not an approach to building software people should use. At least not yet.»
«We should also remember that current token pricing is almost certainly subsidized. These patterns may not be economically viable for long. And those discounted coding plans we’re all on? They might not last either.»
« I ended up building and building and creating a ton of tools I did not end up using much. “You can just do things” was what was on my mind all the time but it took quite a bit longer to realize that just because you can, you might not want to.»
lucumr.pocoo.org/2026/1/18/ag...
Takeaway 3a: Experienced developers find agents suitable for accelerating straightforward, repetitive, and scaffolding tasks if prompted with well-defined plans. Beyond writing new code and prototyping, these suitable tasks include writing tests, documentation, general refactoring and simple debugging. Takeaway 3b: But, as task complexity increases, agent suitability decreases. Experi- enced developers find agents unsuitable for tasks requiring domain knowledge such as business logic, and no respondent said agents could replace human decision making, in part because the generated code is not perfect on the first shot. Takeaway 3c: Experienced developers disagree about using agents for software planning and design. Some avoided agents out of concern over the importance of design, while others embraced back-and-forth design with an AI.
These are the most relevant takeaways:
«Experienced developers find agents suitable for accelerating straightforward, repetitive, and scaffolding tasks if prompted with well-defined plans. But, as task complexity increases, agent suitability decreases.»
«Exp devs retain their agency out of insistence on fundamental software quality attributes, employing strategies for controlling agent behavior leveraging their expertise. They feel positive about incorporating agents into development given their confidence in complementing the agents' limitations.»
Great paper just published, analyzing field observations and surveys from August to October to understand how professional developers use AI.
«Professional developers do not vibe code. Instead, they carefully control the agents through planning and supervision»
arxiv.org/abs/2512.14012
The history behind the technology that allows integrated chips in the scale that are produced today is quite fascinating. It’s amazing to see all the different challenges that were overtaken to produce this.
youtu.be/MiUHjLxm3V0
Great questions to make at the end of your interviews to reveal how the company your are applying really is.
«An interview isn’t just an audition for you; it is a due diligence process for the company.»
dollardhingra.substack.com/p/questions-...
Enshittification happened also to games.
Usually this concept is applied only to websites or other online services, but exactly the same patterns have made games worse over the years.
pleromanonx86.wordpress.com/2025/05/06/w...
«AI changes how developers work rather than eliminating the need for their judgment. The complexity remains. Someone must understand the business problem, evaluate whether the generated code solves it correctly, consider security,ensure it integrates properly, and maintain it as requirements evolve»
«Each advancement addressed a real friction point. Yet the fundamental challenge persists because it’s not mechanical. It’s intellectual. Software development is thinking made tangible. The artifacts we create are the visible outcome of invisible reasoning about complexity.»
«The accumulated decisions, edge cases, and interactions create genuine complexity that no tool or lang can eliminate. Someone must think through these scenarios. That thinking is software development, regardless of whether it’s expressed in COBOL, a CASE tool diagram, Visual Basic, or an AI prompt»
«The challenge is that software development isn’t primarily constrained by typing speed or syntax knowledge. It’s constrained by the thinking required to handle complexity well.»
www.caimito.net/en/blog/2025...
Meta has removed or restricted dozens of accounts belonging to abortion access providers, queer groups and reproductive health organisations in the past weeks in what Repro Uncensored call one of the “biggest waves of censorship” on its platforms in years.
And there is much to be optimized:
Last year I compared Embrace with Datadog RUM, submitting same custom telemetry to both (but each instruments different default events):
In a 5min session browsing a few screens:
- Embrace: 16 requests, 673 kb in total
- DatadogRUM: 60 requests, 1.8 mb in total.
Screenshot of the JSON request payload showing two different events where several contextual data is repeated: Version, session id, OS details, device details, etc.
I just inspected Datadog RUM traffic in my android app and this is not how they are doing. Same “contextual" data is repeated over and over in each event.
I am sure what you are rightly proposing is not what everyone is doing in the observability world...
Screenshot from Spotify app showing Albums from Henry Purcell. There is a separated option to search for “Compilations”, but the selected one is “albums”. Nevertheless the 20 top albums displayed in the screenshot are compilations such as “Autumn Sonatas”, “Sonorous Mexico”, “Back to School - Classics”, “Baroque Masters”… or even “A Classical Melody: Bach & Friends”. Not a single album with complete workpieces from the selected compositor Henry Purcell.
Why #spotify hates classic music listeners?
It’s a pain to find albums from one compositor.
Everything are compilations where you can only find isolated movements mixed with other compositors you are not interested. Quite hard to find complete work pieces.
And compilations is not even selected!
«If you foster an engineering culture where developers are made to feel that management values fast and cheap over continuous progress and quality software, that will have a tremendously negative impact on both the quality and the timely delivery of your software.»
«To cut costs, engineering managers often rush developers, impose arbitrary unrealistic deadlines, or outsource engineering to cheap contractors to try to increase production bandwidth.»
«An investment in quality is an investment in productivity, cost savings, and stronger sales.»
«We invest heavily in software quality because it helps us move faster and save money in the long-run.»
medium.com/javascript-s...
«But if "technical skills" are the skills we use to produce our work (good software) then by extension, every field has "technical" skills. They're simply the skills used to produce the work.»
sashalaundy.com/writing/tech...
But:
«Delays don't always equate to time wasted. Developers typically don't sit idly by while waiting for a code reviews —instead, they go do other productive work»
«Slow build times cause developers to context switch more often, which has negative consequences on productivity not easily measured»
«You've got to be able to translate technical issues to a language that non-technical people understand. And in business, the common language is money. When you can translate technical issues into dollar values, it helps.»
abinoda.com/devex-dollars
«In a tech industry that mistakes velocity for productivity, no one is incentivized to slow down and think about the emergent properties of all those features. No matter how delightful they are individually, the aggregate of all that delight is misery.»
productpicnic.beehiiv.com/p/ux-governa...
Engineering goes beyond what you do when you’re talking to computer systems; it’s also about how you talk to humans. So sometimes being a good engineer boils down to being a good colleague. If you’re mature, constructive, and accountable, you’re telling your new grads that’s what a senior engineer does. If you’re condescending, impossible to please, or never available, that’s what a senior engineer does, too. You shape your company every day, just by how you behave.
Optimize for maintenance, not creation Software is created once, but it will need to be maintained for years. If you’ve got a binary running in production, it will need monitoring, logging, business continuity, scaling, and so on. Even if you intend to never touch the code again, the technical or regulatory ecosystem may force you to care: think of all the old systems that needed to be updated for Y2K, to support IPv6 or HTTPS, or for compliance concerns like SOX, GDPR, or HIPAA. Those won’t be our last disruptive changes. (2038 is coming!) Software gets maintained for much longer than it takes to create it, so don’t build code that’s hard to maintain. Here are some ways you can help Future You and your future team.
Create institutional memory Every time someone leaves your company, you lose institutional knowledge. If you’re lucky, you have some old-timers storing history in their brains. But eventually, inevitably, you’ll have complete staff turnover. When an old system breaks, there’ll be nobody left to say “Oh, yes, I remember when we ran into this before. Here’s what we did last time.” My ex-colleague John Reese, at the time a principal engineer at Google, often also took the role of systems historian: he curated a record of how the site reliability organization had evolved and how running software in production had changed over the years. To create institutional memory, he wrote in-depth articles about the parts of the ecosystem he knew best, then interviewed others to uncover the past, documenting formative systems and practices. Although he’s moved on from Google now, that history lives on with a new set of curators.
What Does It Mean to Do a Good Job? Most of all, you’ll be a role model. How you behave is how others will behave. You’ll be the voice of reason, the “adult in the room.” There will be times when you’ll think “This is a problem and someone should say something”...and realize with a sinking feeling that that someone is you. When you model the correct behavior, you’re showing your less experienced colleagues how to be a good engineer. Later, in Chapter 8, we’ll look at how to actively, deliberately influence your organization and colleagues for the better. But this chapter is about passive influence, the kind that you have just by the way you act as an engineer and as a person.
«Eng goes beyond what you do when you’re talking to computer systems; it’s also about how you talk to humans»
«Software gets maintained for much longer than it takes to create it, so don’t build code that’s hard to maintain»
«Every time someone leaves your company,you lose institutional knowledge»
«Most of all, you’ll be a role model. How you behave is how others will behave. You’ll be the voice of reason, the “adult in the room.”»
newsletter.pragmaticengineer.com/p/the-staff-...
El marco FIC (Frequent Intentional Compaction) Como explica Dex Horthy en su charla sobre Context Engineering, el problema de contexto se resuelve con un flujo de trabajo estructurado: “The key is to separate research, planning, and implementation into distinct phases with frequent intentional compaction.” El marco FIC propone 4 fases independientes: Research: Investigación sin implementación Plan: Diseño iterativo antes de código Implement: Ejecución por fases Validate: Verificación sistemática Entre cada fase, se hace limpieza intencional del contexto (/clear), pero el conocimiento persiste en archivos estructurados. Stepwise-dev automatiza este flujo de trabajo FIC, proporcionando comandos específicos para cada fase y gestionando automáticamente la persistencia del conocimiento en el directorio thoughts/.
Muy prometedor este modelo de trabajo para conseguir que tu agente LLM funcione de forma competente.
«Las pruebas demuestran que después del 50-60% de la ventana de contexto, la precisión cae entre 20-50% dependiendo del modelo.»
nikeyes.github.io/tu-claude-md...
«El valor del programador, per se, baja. La IA ayuda a cualquiera a producir código de calidad.
El valor del experto sube… pero hay menos expertos (Staff+ engineers). Cuando todo el mundo produce más, los errores y la complejidad crecen exponencialmente.»
enespanol.joaoqalves.net/p/cuando-el-...
