Calculette
Empreinte environnementale de Datacenters
Par hubblo
Sometimes I feel that web development has gotten pretty complicated. Fairly straightforward applications are split apart with their frontends powered by behemoth frameworks like React, Next, or Nuxt.
But, it doesn't have to be that way. In this video, I show how a basic backend framework (in this case, Laravel) with its templating language, can have full-stack applications built around the engine of hypermedia, the actual HTML in a browser. It's a strange way of looking at development, but has the potential to remove a lot of headaches, letting the modern standards present in HTML handle the workload, being enhanced at times with a sprinkling of vanilla JavaScript.
The result is an accessible frontend that's fast and scalable, with your data and state built-in.
Some helpful links: https://htmx.org https://alpine-ajax.js.org https://unpoly.com
Welcome! I am going to pass on to you the secret to a successful and brief noobitude, and I won’t even keep you in suspense: nobody cares how many tasks you complete. Why not, and what we care about instead are the subject of the rest of this essay.
Look at your situation from our perspective (by “our” I mean “older engineers”). We hired a bunch of people like you. Some of y’all (we’ll call them A’s) will be amazing game-changers, making everyone around them wildly more productive. Many of you (B’s) will be solid performers. Some of you (the C’s) won’t be here in a year.
We seniors have our regular work to do, but we also have to figure out which category you fit into. We support the superior performers as much as we possibly can. We support the solid performers enough to help them mature. Brutal as it seems, we’d like to expend as little effort as possible on people who aren’t going to make it.
It’s your job to get in the category you want to be in and send us the signals that tell us that’s where you belong.
That stack of tasks you have to do? Your manager or your tech lead could finish those in much less time and with much less hassle than it takes to help you through them. If all we cared about was today’s productivity, we wouldn’t have hired you at all. Instead, we (the seniors) are focused on the future: we know there’s going to be far more work here than we could possibly accomplish. We are paying your salary now as the option premium on the engineer you are going to become. If we play this game right, we’ll have a kick-ass next generation of engineers. If not, we’ll have to be doing the same engineering jobs ten years from now, and we really don’t want to be doing that.
m a huge fan of (the potential for) #LLMs and how they might revolutionize the nature of work and a huge critic of the trillion-dollar shill and #genAI-as-junior-developer / excuse for firing half your workforce that you over-hired over the last decade so really it is you who should be clearing your desk.
I am disgusted by the consequence-free wholesale IP theft and casual planet burning and excited by the potential for local models to do 99% of the heavy lifting.
I am happily speaking to various Claude models in a conversational way and painfully aware that it is in no way 'alive' or 'conscious' and merely acting as a mouthpiece for thousands of people's brilliant work that I am able to lean on.
At the same time.
Why is this so confusing for people? Perhaps we don't need polarized shrieking. Perhaps it is possible for people to contain multitudes.
Categorize each piece of code and then define conventions for each category, using Oxlint for enforcement.
A Series C fintech with ninety engineers, AI-native from day one, lost $4.2 million to a reconciliation function nobody on the team could explain. This is the labor-market thesis on what happens next: why technical autonomy — the ability to ship without a chat window — is becoming the new luxury status in software engineering, and why the engineers who never stopped knowing how the cement works are about to be the most expensive hires of the decade.
We unpack the concept of Comprehension Debt — the silent liability accumulating in every codebase where models write code and humans approve it without understanding it — and trace its consequences across three vectors: the broken senior-engineer pipeline, the bifurcating consulting market, and the compliance frameworks (SOC 2, PCI-DSS, HIPAA, ISO 27001, the EU AI Act, DORA) that refuse to accept "the assistant suggested it" as a control.
The post-AI developer is not the engineer who refuses AI. It is the engineer who can work as if AI did not exist.
━━━━━━━━━━━━━━━━━━ RELATED VIDEO ON THE CHANNEL ━━━━━━━━━━━━━━━━━━
If you want the actual rates, retainer structures, and the specific subfields where the post-AI premium is concentrating right now, watch:
- No-AI Developers Are Billing $250 An Hour — Here's the Market. The numbers are not theoretical. They are this quarter.
━━━━━━━━━━━━━━━━━━ CHAPTERS ━━━━━━━━━━━━━━━━━━
- 0:00 The Paradox of Ease
- 0:44 The Suffering Was the Curriculum
- 1:44 Comprehension Debt
- 2:37 Why Prompt Engineers Are Cheap
- 3:24 Compliance Cannot Terminate in a Model
- 4:19 The Elite Market Is Already Here
- 5:20 How To Be a Post-AI Developer
- 6:04 Cut Your Internet
- 6:51 The Cement Holds
━━━━━━━━━━━━━━━━━━ KEY CONCEPTS ━━━━━━━━━━━━━━━━━━
- The Post-AI Developer — an engineer who uses AI but can ship without it. Defined by autonomy, not abstinence.
- Comprehension Debt — the accumulating liability of merged code that no human on the team can explain. Compounds silently. Surfaces in production.
- The Junior Gap — the broken forging path. Juniors can ship in an afternoon what used to take three weeks. The map of how systems actually work is no longer being built.
- The Bifurcating Market — infinite supply of prompt-driven code generators on one side; vanishing supply of engineers who can debug a heap dump on an air-gapped server on the other.
- The Compliance Wedge — regulated industries (banking, healthcare, defense, payments) require a named human in the chain of responsibility. Language models cannot sign an audit.
- Technical Autonomy — the ability to think without a chat window. The new luxury status in engineering.
━━━━━━━━━━━━━━━━━━ FREQUENTLY ASKED ━━━━━━━━━━━━━━━━━━
What is a post-AI developer?
A post-AI developer is an engineer who uses AI tools but can work, debug, and ship without them. The defining test: if your internet were cut right now, could you still deliver? Post-AI developers say yes. Wrappers around a language model say no.
What is comprehension debt in software engineering?
Comprehension debt is the liability that accumulates when AI-generated code is merged into a codebase without any human on the team being able to explain why the code is structured the way it is. It compounds silently — the function works, the tests pass, until six months later it fails in production and nobody can reason about the original logic. Distinct from technical debt: technical debt is code you understand and chose not to clean up. Comprehension debt is code you shipped without ever understanding.
Why are engineers who don't use AI suddenly billing more than those who do?
Two forces. First, scarcity: the supply of engineers who can debug from first principles — without an internet connection, without a chat window — is shrinking while demand for incident response stays constant. Second, criticality: regulated industries (SOC 2, PCI-DSS, HIPAA, EU AI Act, DORA) cannot terminate the chain of responsibility in a language model. Scarcity plus criticality equals premium pricing. The market is already paying it.
Is being a post-AI developer the same as refusing to use AI?
No. Refusing AI is what the video calls "Luddite cosplay" — the market does not pay for performative purity. Being post-AI means using AI freely while remaining able to defend every line, notice every silent error, and continue working when the assistant is unavailable. It is a stricter standard than refusal, not a softer one.
How do junior developers become senior engineers in a post-AI world?
By deliberately rebuilding the map that AI removed. That means doing some work — debugging, reading source, reasoning about memory and network behavior — without the assistant, even when using it would be faster. The friction is the curriculum. Engineers who skip it can ship features but cannot own incidents, and the labor market is starting to price that distinction explicitly.
In 2026, anyone with a terminal can type eight lines of telnet and send mail claiming to be your CEO. No password. No certificate. No proof of identity.
The protocol that carries every email on the public internet was specified in August 1982 for 235 trusting research hosts — and nobody has re-shipped it since.
This is the SMTP Tax: the cumulative cost of running a 44-year-old plaintext mail protocol on a network that now moves roughly 320 billion messages a day between adversaries. Spam filters, BEC wire fraud, midnight SPF flattening, multi-billion-dollar email security industries — all of it is the bill for an architectural default the world stopped questioning around 1995.
The video walks through the full stack of patches that have been bolted onto SMTP since 2003. SPF tells receivers which IPs are allowed to send for a domain — and breaks the moment a message gets forwarded. DKIM signs message bodies with real public-key cryptography — and says nothing about the visible From: header your user actually reads. DMARC ties the two to the visible identity through alignment — and is advisory, with the most common policy being p=none, which rejects nothing. BIMI publishes a logo for domains that already enforce DMARC — a UX sticker on top of three optional patches. MTA-STS, ARC, TLS-RPT, DANE — each one a layer on a layer on a layer.
Then the structural argument: what email looks like if you stop bolting. JMAP (RFC 8620) moves authentication from a TXT record to the request itself. The decentralized identity world — DIDs, the kind of identity model behind Bluesky's AT Protocol — asks what mail looks like if every sender is a key, not a string. The reason that argument is approximately as far away as IPv6 was twenty years ago isn't technical. It's gravitational.
The closing frame: the SMTP Tax is not a bug. It's a default chosen in 1982 for a federation of universities, never re-chosen for the network that exists now.
━━━━━━━━━━━━━━━━━━ CHAPTERS ━━━━━━━━━━━━━━━━━━
- 0:00 The Cold Open: Eight Lines, No Password
- 0:19 By The Numbers: 320B Emails, $50B BEC
- 1:20 RFC 821, August 1982: A Network Of 235 Hosts
- 3:11 SPF (2003): The First Patch
- 3:49 DKIM (2007): Strong Primitive, Wrong Target
- 4:28 DMARC (2012): Why p=none Is Theatre
- 5:20 BIMI And The Patch Pile
- 6:17 The Structural Alternative: JMAP
- 7:01 Cryptographic Sender Identity And DIDs
- 7:59 A Default Nobody Re-Chose
━━━━━━━━━━━━━━━━━━ KEY CONCEPTS MENTIONED ━━━━━━━━━━━━━━━━━━
- SMTP — Simple Mail Transfer Protocol, RFC 821 (1982), Jon Postel
- RFC 793 — TCP specification, source of Postel's Law
- Postel's Law — "Be conservative in what you do, be liberal in what you accept from others"
- SPF — Sender Policy Framework (2003)
- DKIM — DomainKeys Identified Mail (2007)
- DMARC — Domain-based Message Authentication, Reporting and Conformance (2012)
- BIMI — Brand Indicators for Message Identification (2020)
- MTA-STS, ARC, TLS-RPT, DANE — auxiliary patches on the SMTP stack
- JMAP — JSON Meta Application Protocol, RFC 8620 (2019)
- BEC — Business Email Compromise (FBI IC3 category since 2013)
- DID — Decentralized Identifier (used by Bluesky's AT Protocol)
- ARPANET — 235 hosts at the time RFC 821 was written
━━━━━━━━━━━━━━━━━━ FREQUENTLY ASKED QUESTIONS ━━━━━━━━━━━━━━━━━━
What is the SMTP Tax?
The cumulative cost of running a mail protocol designed in 1982 for 235 trusting hosts on a 2026 network of 320 billion daily messages between adversaries. It includes BEC fraud, spam filtering infrastructure, the entire email security vendor industry, and the engineering hours spent maintaining SPF, DKIM, DMARC, and BIMI records that the protocol itself never required.
Why doesn't SMTP have built-in sender verification?
Because RFC 821, written by Jon Postel in August 1982, specified SMTP for ARPANET — roughly 235 hosts where every administrator knew every other administrator by name or institution. The threat model was hardware failure, not a hostile actor. The word "authentication" does not appear once in the 47-page document.
Does DMARC stop email spoofing?
Only when a domain publishes p=reject and the receiver honors it. The most common DMARC policy by an enormous margin is p=none, which is monitoring mode — it rejects nothing. Roughly 30% of Fortune 500 domains enforce p=reject; across the broader Alexa top 1 million, the figure is closer to 12%. Everyone else is running a record that exists so a vendor checkbox lights up green.
Modern software is 43x slower than it was twenty years ago, and the people building it cannot tell you why. This is the bill for forty years of "hardware will catch up." For two decades the industry traded performance for convenience and called it progress.
Wirth's Law became a roadmap. Dennard scaling died in 2005 and nobody on the team noticed.
A "Hello World" now ships with 1,847 dependencies, allocates 487 MB of memory, and takes three seconds to draw a window.
The de-bloating is what happens when the stopwatch finally wins the argument: Rust at Cloudflare, Zig under Bun, Mojo under the Python data stack, Go quietly eating the JVM.
The layer count is the bug. This video is the diagnosis.
━━━━━━━━━━━━━━━━━━ RELATED VIDEO ━━━━━━━━━━━━━━━━━━
If this was the diagnosis, the next one is the receipt. Watch it after this one: "Stop Buying RAM. Read This Instead. (The Tab Tax)"
━━━━━━━━━━━━━━━━━━ CHAPTERS ━━━━━━━━━━━━━━━━━━
- 0:00 Two Apps, Twenty Years Apart
- 0:35 The Abstraction Trap
- 2:19 The Benchmark Doesn't Lie
- 3:27 The Low-Level Mindset Returns
- 5:14 The Bill Is Due
━━━━━━━━━━━━━━━━━━ KEY CONCEPTS COVERED ━━━━━━━━━━━━━━━━━━
- Wirth's Law (Niklaus Wirth, 1995): software gets slower faster than hardware gets faster
- Dennard scaling and the end of single-thread CPU performance growth around 2005
- The abstraction trap: why component libraries on top of design systems on top of runtimes on top of browsers stopped being free
- The Chromium runtime tax and why excellent engineers inside Electron lose to the stopwatch
- "Clean Code" as superstition: when a style guide becomes a religion at the cost of an order of magnitude
- The low-level mindset returning as economics, not nostalgia
━━━━━━━━━━━━━━━━━━ TOOLS AND LANGUAGES MENTIONED ━━━━━━━━━━━━━━━━━━
- Electron and the Chromium runtime tax
- Rust, Cloudflare Pingora (one trillion requests a month), Discord read-states, the Windows kernel rewrite
- Zig and the Bun runtime
- Mojo and the Python data-science stack (NumPy, CUDA)
- Go: the quiet boring default that ate the JVM
- hyperfine for cold-start benchmarking
- npm and the 1,847-dependency baseline
━━━━━━━━━━━━━━━━━━ FREQUENTLY ASKED QUESTIONS ━━━━━━━━━━━━━━━━━━
What is Wirth's Law?
Wirth's Law, formulated by Niklaus Wirth in 1995, states that software gets slower faster than hardware gets faster. It was meant as a warning. The industry treated it as a roadmap for forty years.
Why is Electron so slow and memory-hungry?
Electron ships a full Chromium runtime with every app, which means each "native" desktop app is paying the cost of an entire web browser before the application code even runs. In the video, a Rust-based editor opens a 1 MB file 43 times faster and uses 22 times less memory than its Electron equivalent on the same machine.
Is JavaScript or the web stack dying?
No, but the assumption that hardware would keep absorbing software bloat for free is dying. Dennard scaling ended around 2005 and single-thread CPU performance stopped doubling. The "de-bloating" is the industry quietly rebuilding hot paths in Rust, Zig, Mojo, and Go because the layer count became the bottleneck.
What does "the layer count is the bug" mean?
It is the recognition that modern application slowness is not caused by any single bad framework, but by accumulating runtimes on top of runtimes until the abstraction tower itself becomes the dominant cost. Peeling layers, not adding them, is now the optimization.
What should I watch next? "Stop Buying RAM. Read This Instead. (The Tab Tax)" — same channel.
That video is the receipt: the most ordinary place on your machine where the bloat bill arrives first, and the one most people already opened today and forgot about. This one is the diagnosis. Watch them in that order.
E-commerce, billeterie FOSS
Ils organisent les premières "Forgéales" : des soirées de partage en visio les 24, 25 et 26 juin.
Je découvre une initiative qui s'appelle "La forge des communs numériques éducatifs".
La Forge des communs numériques éducatifs ?
- C’est une communauté d’enseignantes et d’enseignants qui créent et partagent des logiciels et des ressources éducatives libres.
- Ces outils sont conçus pour être utiles à leurs collègues et à leurs élèves, dans leur pratique quotidienne.
- Tout cela se passe dans un espace de travail collaboratif en ligne, qui rassemble plusieurs milliers de projets.
- Chacun peut utiliser ces ressources, les adapter à ses besoins, et surtout, y contribuer librement.
- Un lieu ouvert, vivant, fait par et pour la communauté éducative !
Ils organisent les premières "Forgéales" : des soirées de partage en visio les 24, 25 et 26 juin.
Query engines make me feel like a wizard. I cast my incantation: “Give me all the directors and the movies where Arnold Schwarzenegger was a cast member”. Then charges zip through wires, algorithms churn on CPUs, and voila, an answer bubbles up.
How do they work? In this essay, we will build a query engine from scratch and find out. In 100 lines of Javascript, we’ll supports joins, indexes, and find our answer for Arnold! Let’s get into it.
Suivez la table ronde réunissant des acteurs du monde du logiciel libre :
- Renaud Chaput, directeur technique de Mastodon ;
- Loïc Dayot, membre du conseil d’administration de l’April, et Étienne Gonnu, chargé de mission affaires publiques ;
- Pierre-Yves Gosset, coordinateur des services numériques de Framasoft ;
- Nicolas Vivant, fondateur de France Numérique libre, directeur de la stratégie et de la culture numériques de la commune d’Echirolles.
Le cas de la ville d'Echirolles
En fin de mandat, donc en 2026, nos objectifs sont atteints.
Microsoft, Google, Amazon, Oracle ; nous n'avons plus aucune dépendance envers les grands acteurs de la tech. Des événements tels que le rachat de VMware par Broadcom, à Windows 11, ou les différentes pannes récentes ne nous ont en aucune façon affecté.
Nous maîtrisons non seulement notre infrastructure et nos outils, mais aussi notre budget. Les économies en coût de fonctionnement sur la durée du mandat s'élèvent à plus de 2 millions d'euros.
Nicolas Vivant, fondateur de France Numérique libre, directeur de la stratégie et de la culture numériques de la commune d’Echirolles.
— audition à l'assemblée nationale du 6 mai 2026