diff --git a/skglm/tests/my_test.py b/skglm/tests/my_test.py
deleted file mode 100644
index 9efadf0d..00000000
--- a/skglm/tests/my_test.py
+++ /dev/null
@@ -1,39 +0,0 @@
-# %%
-import numpy as np
-from skglm.datafits import HessianQuadratic, Quadratic
-from skglm.solvers import AndersonCD
-from skglm.penalties import L1
-from skglm.estimators import L1PenalizedQP, Lasso
-import scipy as sp
-# %%
-n = 100
-d = 3
-G = np.random.randn(n, d)
-A = G.T @ G + np.eye(d)
-b = np.random.randn(d)
-U, s, Vh = sp.linalg.svd(A, full_matrices=False)
-S = np.diag(s)
-A_half = U @ np.diag(s**(0.5)) @ Vh
-A_minus_half = U @ np.diag(s**(-0.5)) @ Vh
-X = np.sqrt(d)*A_half
-y = -np.sqrt(d)*A_minus_half @ b
-# %%
-x = np.random.randn(d)
-Ax = A @ x
-hessian_quadratic_loss = HessianQuadratic()
-hessian_quadratic_loss.value(b, x, Ax)
-# %%
-Xw = X @ x
-quadratic_loss = Quadratic()
-quadratic_loss.value(y, None, Xw) - (1.0/(2*d))*(y**2).sum()
-
-# %%
-alpha = 1e-1
-qp_solver = L1PenalizedQP(alpha=alpha, fit_intercept=False, max_iter=500)
-lasso = Lasso(alpha=alpha, fit_intercept=False, max_iter=500)
-
-qp_solver.fit(A, b)
-lasso.fit(X, y)
-# %%
-print(f'{qp_solver.coef_=}')
-print(f'{lasso.coef_=}')