Paul ARGOUD

DuckDB : your RAG fits in a file, not a cluster

Artificial Intelligence

A RAG PoC indexing 84,000 chunks, roughly 500 MB of vectors, fits without flinching in the RAM of an M1 MacBook; and yet French SMEs are still being sold a four-service stack (Pinecone + Elasticsearch + Postgres + LangChain) that costs between €150 and €400 a month in cloud infrastructure and mobilizes three distinct technical skill sets. DuckDB, the embedded analytical library maintained since 2019 by Amsterdam’s CWI, solves the equation with a single dependency: native vector search (the vss extension), BM25 full-text search (the fts extension), standard structured SQL filters, and hybrid re-ranking in one eight-line query, all of it inside a local .duckdb file. The bottom line: a factor of 10 on infrastructure cost, a factor of 10 on latency, a factor of 5 on time to production, and not a single embedding leaving your infrastructure to end up in an American datacenter. The four-service stack taught today in 90% of RAG tutorials doesn’t exist to solve a technical problem: it exists to sell recurring SaaS, to justify DevOps engagements, and to inflate project budgets on use cases that don’t warrant it. The real value of a senior engineer in 2026 is measured by the number of lines of infrastructure he knows how to remove to solve the exact same problem, not by the number of Kubernetes microservices he knows how to stack.

Your RAG is lying to you : why your eval metrics are (probably) wrong

Artificial Intelligence

I changed my chunking from 800 to 1,200 tokens, my hit@1 jumped +12%, and for thirty seconds I thought I’d found the optimum. In reality, my questions about the MSS60 were now being evaluated against the MSS54 index. The routing between corpora had been broken from the start, without anything in the metrics flagging it. A rising metric produced by a broken eval pipeline looks exactly like a rising metric produced by a genuine improvement. Three silent traps, multi-corpus routing, the absence of versioned history, and hit@k and MRR with no LLM judge, mean that most RAG teams make their decisions on numbers that measure nothing.

Real SVG, fake SVG : Recraft V4 and the end of the vector designer

Artificial Intelligence

Most AI SVG generators cheat: they hand you a base64-encoded PNG dressed up with a .svg extension. The test takes ten seconds, open the file in a text editor. In 2026, Recraft V4 is the only consumer tool that reliably produces real native SVG from a prompt, with Adobe Firefly Text-to-Vector as a legally better-protected alternative. Three trades collapse along the way: junior UI icon designer, low-end freelance logotype, vectorizer. What gains value is the upstream (brand design, design-system architecture) and the downstream, the technical integration.

Structured products : you’re not investing, you’re insuring your bank against crashes

Economy

On 6 February 2026, Stellantis shares lost 25% in a single session, taking with them an illusion carefully maintained for twenty years: these savers had not bought a “protected” investment; they had sold their bank a catastrophe insurance policy disguised as a coupon. Some buy-back valuations now show -99% on instruments stamped “with safety barrier”, and yet the French structured products market has tripled in four years to reach €60 billion in net inflows in 2025, because no other product generates as much revenue for banks. The mechanics are crystal clear once you strip away the marketing: you surrender the dividends, you cap your gains at the coupon, and you retain 100% of the downside risk beyond a barrier whose statistical robustness amounts to sophistry, while one in five CAC 40 stocks has experienced a drawdown exceeding -50% over the past twenty years. Belgium settled the matter as early as 2011 with an FSMA moratorium that eliminated single-stock barrier autocalls from the retail market, without any collapse of the savings market; France, meanwhile, is still publishing AMF/ACPR mapping reports calling for “more financial education”. At some point, we will have to stop pretending.

Levi’s : chaos as a business strategy

Economy

I have been loyal to Levi’s for decades, and yet I have never owned the same pair of jeans twice. Every time I replace a pair, something has changed: cut, colour, fabric composition, weight. This is not disorder. It’s a strategy. By making product loyalty impossible, the brand exploits precisely the loyalty to the icon: the red tab, the Far West mythology, a carefully maintained heritage designed to conceal a denim that has gone from 14 to 15 oz in the 1980s to 10 to 12 oz today. What is the point of being loyal to a brand that makes loyalty to its product impossible?

Anthropic’s HTML manifesto : a tax on intelligence disguised as progress

Artificial Intelligence

On 8 May 2026, Thariq Shihipar, an engineer on the Claude Code team at Anthropic, published a manifesto that’s been circulating ever since: drop markdown, switch to HTML for your specs, implementation plans and reports. The piece is clever, well written, and structurally an act of strategic marketing dressed up as personal experience. On a narrow subset (throwaway editors, interactive playgrounds, prototypes with sliders), the author is right; on everything else, his blanket enthusiasm conceals six blind spots. Token cost waved away, semantic HTML flipped into presentational HTML, loss of reviewability, the maintenance paradox that traps the user inside the production loop, reading that gets skimmed rather than scrutinised, an attack surface opened up by indirect prompt injection. An analysis of a poorly disguised strategic signal, plus a pragmatic decision matrix for sorting out where HTML deserves to be adopted and where markdown remains, by design, the right pivot format.

Claude Code is amnesiac. Understand-Anything cures it.

Artificial Intelligence

Every session, Claude Code reopens your codebase like a visitor who has never set foot in it. It unfolds the README, fires off its greps, opens whole files to reconstruct an architecture it had already pieced together the day before. Understand-Anything, an open-source project released in March 2026, offers another path: index the codebase as a local knowledge graph, versioned like a lock-file, queried on demand by the agent instead of being re-read on every turn. This isn’t a token optimisation, it’s an architectural shift that extends the “RAG is dead, long live the Agent” thesis right into the code itself. Multi-agent pipeline, comparative economics, the limits of a project just six months old: what this break changes for Claude Code and for your workflow.

wp_options : archaeology of a technical debt with no owner

Softwares

For twenty years, WordPress has let every plugin name its options and tables however it likes. The add_option() function sets autoload = ‘yes’ by default, with no warning, no documentation urging caution. The register_uninstall_hook() hook exists but is mandatory for no one. The uninstall.php file is optional. And the API provides no discovery mechanism letting the administrator find out, after the fact, which plugin a given option or table belonged to.

France at a standstill : neither fast nor at a reasonable cost

Democracy

Three days without bank transfers, two years to renovate a five-million-euro rural bridge, eleven months to repair three escalators at Châtelet, two weeks without internet in Saint-Marcellin. These situations are no longer accidents: they have become the cruising speed of a country that has lost its ability to do things fast and at a reasonable cost. At the root of this paralysis, four structural causes that reinforce one another; cascading subcontracting, administrative layering, the asymmetry of privatisation, devouring bureaucracy. A million euros has become the minimum unit at municipal scale, a billion at national scale, and the citizen ends up taking out a Starlink subscription to get what the State, the operators and the local authorities no longer know how to guarantee. A methodical slide and a way out, on condition of shortening decision chains and rehabilitating local doing.

If the profits belong to the workers, so do the losses, no ?

Economy

The refrain is familiar: when the firm makes 200,000 euros in profit, it’s the workers’ labour that created the wealth. But when it loses 50,000 euros because the market turned, far fewer voices call for symmetry. The employee risks their job, their unpaid back wages, their pension. The shareholder risks their stake. Both take a risk, but not the same one, and it’s precisely this imbalance that has to be discussed honestly, figures in hand. A coherent answer does exist, by the way: the cooperative, where everyone is exposed. And when it breaks, as at Duralex, it’s the worker-shareholders who lose their savings. So, the real question: do you accept the symmetry, or do you just want the upsides of capitalism without the downsides?