Announcement from **Meta AI Research** published on **August 5, 2026** (stated reading time: 4 minutes, no individual byline): **Muse Code** in beta, *"a terminal coding agent"*, and the model that powers it, **Muse Spark 1.2**. Meta itself frames the launch: *"This marks our next step toward the frontier, with larger and much more capable models on the way"* — **a step toward the frontier, not a big claim**. **Three architectural elements** are described on the harness side: **asynchronous background agents** that *"remain active throughout each session, rather than being spawned for individual tasks"*, to avoid redundant information gathering and reduce the need for steering; a **local event log** where *"every model call, tool run, approval, and edit is appended"*, making the runtime *"replay-exact and restart-safe"* — able to resume exactly where it left off after a crash; and **three skills shipped out of the box** — `/plan` (turns a task into a plan submitted for approval), **`/grill`** (stress-tests the plan *"until it holds up"*), and `/goal`. On the model side, Meta claims **co-training the model with the harness** (*"to maximize harness compatibility"*, with harness trajectories sampled via rejection sampling and recipe optimizations for goals, compaction, and sub-agents), **long-horizon** training (whole-repository generation, end-to-end projects, self-research, with planning, goal conditioning, and **context compaction**), and a **self-improvement loop** in which Muse Spark 1.1 generates the environments and instruction templates **and then grades candidate solutions**, producing a training set for 1.2. ⚠️ **The most notable fact in this announcement is written nowhere in its text**: the four published charts — Terminal-Bench 2.1, DeepSWE 1.1, an internal Meta benchmark, and the GPU kernel optimization case study — **place Muse Spark 1.2 behind Opus 5 in all four cases**, including on Meta's own proprietary benchmark (70.6% vs 79.4%) and on the case study, where the model finishes **fourth out of six** (+68.7% vs +74.0%). ⚠️ **And the model's actual gain is smaller than it appears**: on the two public benchmarks, 1.1 is measured with `mini-swe-agent` and 1.2 with Muse Code — the 6.7-point gap therefore conflates model and harness. On the internal benchmark, the only comparison where no harness is mentioned, the 1.1 → 1.2 gap drops to **2.3 points**.
#Meta AI Research#Muse Code#Muse Spark 1.2
**Meta AI Research** — publication institutionnelle sans auteur nommé · sur `research.meta.ai`. Le billet renvoie à un **rapport** pour la méthodologie d'évaluation · non repris ici.
Tech-watch note by **Didier Girard** dated **August 2, 2026**, prompted by a colleague's question ("what is ACP?") that turns out to address a problem that is not terminological but **documentary**. **Three protocols compete for the acronym**, with no technical overlap whatsoever: **Agent Client Protocol** (client ↔ agent — Zed, August 2025, JSON-RPC 2.0 over stdio, Apache-2.0, "what LSP did for languages"), **Agentic Commerce Protocol** (agent ↔ merchant — OpenAI + Stripe, Sept. 29, 2025, facing Google's **UCP** of Jan. 11, 2026 backed by **AP2**), and **Agent Communication Protocol** (agent ↔ agent — IBM Research / BeeAI, marginal but polluting searches). **The core of the note is not the disentangling but its observed failure**: the author searches "ACP" in his tech-watch knowledge base and gets **twelve results, all about the commerce protocol, zero about Zed's** — *"our watch agents had indexed the acronym without disambiguating it"*. Hence a knowledge-engineering rule: ***"never index a bare acronym"*** — the entity is "Agent Client Protocol", "ACP" is **only an alias**, carried by three distinct entities. A structuring clarification follows (**MCP connects an agent to its tools, ACP connects a client to an agent; the two stack**), then the textbook case: **Buzz**, published by **Block** on July 21, 2026 under Apache-2.0 — a self-hostable workspace built on **Nostr**, where every human or agent participant is a **keypair** and every message, workflow step, or git push is a **signed event** in an append-only log. A fully protocol-based architecture (`buzz-acp` an ACP harness over stdio, `buzz-agent` an ACP agent calling an LLM, `buzz-dev-mcp` an MCP server for shell + editing), hence the agent-agnosticism: **Goose, Claude Code, and Codex** plug in through the same harness, and **Hermes** (Nous Research) connected to it without Block writing a single line — *"N+M instead of N×M, running in production"*. The note closes on the question of the **Claude subscription** versus third-party agents, with a five-stage 2026 timeline and a **design rule** that applies beyond this case: the line is not legal but **architectural** — ***"who is consuming, and on whose behalf"*** (an `owner-only` agent consumes your subscription for you; an `anyone` agent in a shared channel routes your colleagues' requests through your account). ⭐ **Verification carried out on this corpus**: the thesis holds, and more starkly than the note asserts — not only is "Agent Client Protocol" **completely absent** from it, but the bare acronym `ACP` **is already typed as an entity** in two fiches, and the KB page `Agentic-Commerce-Protocol` **already attributes the protocol to Google** when it belongs to OpenAI + Stripe. The collision described is not a risk yet to come: it has **already produced an attribution error** in the graph.
**Didier Girard** — auteur de la note. Écrit ici depuis la position de **praticien de la veille outillée** : le déclencheur est une question de collègue · le matériau principal est le comportement observé de sa propre base de connaissances · et la conclusion est une **règle de curation** adoptée en interne. Le texte alterne donc deux voix — l'explicateur de protocoles et l'ingénieur de la connaissance qui constate un défaut chez lui et en tire une norme.
Episode « Phase 5 · Review » of SFEIR's series on the augmented SDLC, published **the same day** as Addy Osmani's LinkedIn post, which it translates into a phase specification. Thesis: **quality has changed address** — it is no longer read in the code (agents produce more of it than anyone can review) but in **the ring of constraints surrounding the agent**. Osmani's ring (seven dimensions — correctness, security, performance, accessibility, maintainability, **economic efficiency**, **understandability** — linked by the **back-pressure** rule: "a loop is granted only the autonomy one knows how to verify cheaply and reliably, not an inch more") is redrawn, translated, and attached to phase 5 of SFEIR's 11-phase cycle. The structuring corollary: **the bottleneck was never generation, it is verification** — "generation is a wide mouth, verification a narrow neck; speeding up the mouth thickens the pile at the neck." **The most interesting design decision is a cycle-architecture choice**: Review is deliberately **outside the three human gates** (Define, Plan, Ship), because putting the gate on Review would amount to placing human attention — a finite resource — as the checkpoint of a generation capacity that, itself, scales: "you would have built a pipeline whose maximum throughput is the number of diffs a senior can read before the end of the day." Hence the split: **Review instruments, Ship decides** — Review delivers an *opposable body of evidence*, Ship decides on the evidence, not on the full diff. A position taken against Monperrus (whose diagnosis SFEIR keeps — human inspection of every diff does not withstand agentic speed — but whose conclusion it rejects: acceptance cannot be delegated). The named trap is **circular validation** (the agent that writes the code writes the tests that validate it: "you have built a mirror, not a ring"), with five countermeasures borrowed from Anthropic (independent gates in separate context windows, deterministic + agentic never substituting for one another, shadow mode, risk-based tiering, logging to the SIEM) and Compare the Market's warning (**AST graph ~70% vs vector RAG ~58%**, RAG performing *worse than no context at all*). The firm's own extension is **the ratchet**: "every escape becomes a constraint" — a defect that has crossed the ring is closed back up *inside the ring* (test, lint rule, rubric, harness guardrail) at Compound-1, "the only asset in the chain that appreciates while models depreciate" (an internal, unauditable measure: **− 30% fewer correction iterations after ten cycles**). It closes by reformulating the question: "is this code good?" has become unanswerable; what remains is **"what does my system refuse to let through?"**
#ring of constraints#constraints around agents#Review phase
SFEIR (voix éditoriale du cabinet, article non signé individuellement) — construit sur Addy Osmani (Google) ; cite Martin Monperrus · Paula Hingel (Augment Code) · DORA/Google Cloud · Jason Clinton (Anthropic) · l'équipe Engineering de Compare the Market
SFEIR's engineering-cabinet analysis ("an engineer's reading") of the **July 16, 2026** launch of **Kimi K3** by the Chinese laboratory **Moonshot AI**: an **open-weights, frontier-class model** whose provider claims **~2.8 trillion parameters**, a **one-million-token context**, and **weight release before July 27, 2026** (likely under a Modified MIT license, as with the K2 lineage). Thesis: capability once thought reserved for proprietary giants (Anthropic, OpenAI, Google) is becoming available **in open weights, at a discount price, from a Chinese lab**. SFEIR — despite being an **Anthropic and Google Cloud partner**, and thus "with no interest in overselling a Chinese model" — adopts a cardinal **methodological caveat**: on launch day, **no official, complete benchmark table** exists; specs (2.8T, Kimi Delta Attention, +25% training efficiency) and scores are **vendor-stated** or drawn from **community arenas**, "to be treated as claims, not measured facts." The new architecture (**Kimi Delta Attention**, hybrid linear attention; decoding claimed up to **6.3x faster** at 1M tokens) breaks with the K2 cadence (K2 Jul. 2025 → K2.7 Code Jun. 2026, a flagship every two months); two variants accompany the launch (**K3 Max**, **K3 Swarm Max**), with forced sunsetting of the kimi-k2.5/moonshot-v1 series on **August 31, 2026**. **The real weapon is price** (~$3/M input, $0.30 cached, $15 output per secondary sources): a frontier open-weights model at this level **pulls the whole price-performance curve down** — the commoditization of the model layer, accelerated by open source. But the decisive singularity is not a score: it is **reversibility**. A frontier open-weights model turns a consumed API (vendor dependency) into an **option** (self-host, portability, exit from lock-in), at the cost of heavy infrastructure to host 2.8T parameters. SFEIR's view: **open-weights changes the question, not just the answer** — no longer "which model is best/cheapest?" but "how much of my system am I willing to make dependent on a vendor I don't control?". The right posture remains a **routed portfolio** (one model per task, one model per constraint), with Kimi K3 adding a **"reversibility" column** to the decision grid. The "AI Only" conviction stands unchanged: the model is a commodity, the durable advantage lies in the engineering around it (Context Engineering, harness, cost governance, ability to change one's mind). The figures still need validating "on your own" — your repositories, your data.
Op-ed by **Olivier Rafal** (Consulting Director Strategy, **WeNvision** — **SFEIR** group; former editor-in-chief of *Le Monde Informatique*) published on **June 1, 2026** in **CIO-Online**, structured around a **paradox**: in the AI era, software engineering **changes everything… and nothing changes**. **What changes = the operating model.** Roles are redefined: the **Product Owner** shifts from backlog breakdown to **generating context usable by AI**; the **developer** shifts from writing code to **framing, directing, and reviewing** agent execution; **QA** gains the ability to define **expected proof** upfront. Team structure shifts from *"double pizza teams"* (hand-off chains of ~8 people) to ***"sandwich teams"***: a **tight pairing of a business expert and a tech lead, both AI-augmented**, with other skills in support. Internal **Sfeir** figure: *"this pairing now drives roughly 80% of the production chain"*, with the remaining ~20% (architecture, data governance, security) centralized. Pivot quote: ***"The issue is not a tooling issue, but an operating-model issue."*** **What doesn't change = the discipline of the cycle.** The **SDLC** phases (define → build → verify → deploy → maintain) remain identical and non-negotiable; AI removes none of them, it **intensifies** them: ***"all the slack that human-paced work absorbed, one way or another, becomes, at AI speed, industrial-grade defects"*** (amateur-vs-professional sport metaphor). Hence **three inviolable *gates*** (human control): **specification, planning, delivery review**; validation **by proof** (not by AI's own assertions); **systematic capitalization** (each cycle enriches the next) → measured result: **−30% correction iterations after ~10 cycles**. Principle: ***"the faster the execution, the stricter the framework must be."*** Concepts drawn on: **harness** (agentic rules adapted to context), **vibe-coding** deemed **untenable in the enterprise**. **Third pillar = governance, FinOps & value-driven steering**: **variable and recurring** AI costs (~**€10/hour** per augmented seat), a shift from flat-rate licensing to usage-based billing (a 2010s cloud parallel); **FinOps** does not aim to cut costs but to *"optimize tool efficiency"* (cost weighed against value); aligning **business metrics** upfront (time-to-market, features, performance, eco-design). **Conclusion**: acceleration makes the fundamentals **non-negotiable**; the challenge is **organizational and cultural**, not technological — without securing the business relationship and collective discipline, an AI-boosted SDLC merely **amplifies problems** (hitting the wall faster). Extends the WeNvision doctrine from [[rafal-wenvision-ia-generative-produit-techno-pas-projet-2024-02-23]] and [[rafal-wenvision-tokenomics-foundation-finops-ia-2026-06-04]]; converges with *systems around the model* [[dropbox-okumura-beyond-code-generation-engineering-productivity-ai-agents-2026-05-28]], *harness engineering* [[osmani-agent-harness-engineering-2026-04-19]], agentic Salesforce, and the *agent manager* debate (BFM/Girard, SFEIR).
**Olivier Rafal** · *Consulting Director Strategy* chez **WeNvision** (groupe **SFEIR**). Ancien **rédacteur en chef du *Monde Informatique*** · et auparavant consultant analyste du marché IT (~10 ans). Tribune publiée dans la rubrique *Tribune* de **CIO-Online**. Publié le **1er juin 2026**.