LATEST
Stabileo 0.6.0: re-run affected projects
Re-run affected analyses and review the results you already used. SA-2026-048 through SA-2026-063 explain what to check; SA-2026-056 remains unfixed.
Read the postBLOG
Notes on the solver, structural-code checks and the decisions behind them, written by the people who build the software.
LATEST
Re-run affected analyses and review the results you already used. SA-2026-048 through SA-2026-063 explain what to check; SA-2026-056 remains unfixed.
Read the postWhat changed since 0.4.0: twenty-three defects that could give a wrong or unconservative result, written up as advisories and all but one fixed; a warning or a refusal for every defect still open; and the step-by-step course in Basic.
Read the postHow Stabileo tells you which of your projects a fix concerns, how it holds back a result it knows can be wrong until the fix exists, and how the fix is published, with the fourteen results versions merged on 4 and 5 November.
Read the postBasic now has a course: pick a method and the app solves your structure, or an example, step by step as on paper. Statics, influence lines, second order, plastic analysis and dynamics. Free.
Read the postWhat we shipped in the last few days: runnable research, eight calculators, and checks on the benchmark models themselves. What is working, what was wrong, and what still needs review.
Read the postTwo solves of the same model in one session could differ in the last digits. Why that happened, how we fixed it, and how every report now says which version and which data produced it, so anyone can reproduce it.
Read the postA certificate that proves a wrongly assembled system was solved correctly is not worth much. How Stabileo rebuilds every frame model on its own and certifies the result against the model, what that found, and what it leaves out.
Read the postWhat changed since 0.2.0: sixteen defects that could give a wrong result, fixed and published as advisories; certificates that check the result against the model; and the same numbers on every solve.
Read the postOn 20 October we received the second-place certificate of TIIC, the Argentine Chamber of Construction's innovation programme, showed Stabileo at a stand and took part in the talk "Your next site colleague could be an algorithm".
Read the postThe solver, formal verification and AI: how we look for errors in the numbers Stabileo gives, what the certificates prove and what no verification can guarantee.
Read the postThe Stabileo defects that could give wrong or unconservative results: who they affect, how to tell whether a model is affected, what to do in the meantime and when they were fixed.
Read the postFor a prismatic member, the stiffness method is the exact special case of the finite element method. What changes is the model, and for a deep beam the frame model gives a deflection 19 to 34% too small.
Read the postOn 18 September we presented paper 86N in Córdoba: a structural calculation agent that proposes changes and leaves every number to a verified solver. What we showed, why the agent has to be kept apart from the calculation, and what we brought back.
Read the postThree formulas, one section, three different numbers. Which one applies depends on the wall; where two apply at once, Bredt comes out 2 to 15% low.
Read the postAlmost everyone remembers φMn ≥ Mu. That is the last line of the check, not the check. The outcome is decided earlier, and one of those steps can pull φ below 0.9.
Read the postThere is free structural software, and it is good: it models, it solves, it shows results. What we could not find is one that approaches the advanced tools from the conceptual side.
Read the postA language model cannot guarantee that a number is right. In the architecture we propose for structural engineering, the agent interprets and proposes, and every number comes from a verified solver.
Read the post