Solution (source code)

= Solution

Define the <stochastic integral>
$$
\boxed{W_t=\frac1{\sqrt{2\lambda}}\int_1^{e^{2\lambda t}}u^{-1/2}\,dB_u.}
$$
It is adapted to $\mathcal G_t$, starts at zero and has continuous paths. Integrals over disjoint clock intervals are centered Gaussian and independent, and an increment from $s$ to $t$ is independent of $\mathcal G_s$. The <Itô isometry> gives its <variance>
$$
\frac1{2\lambda}\int_{e^{2\lambda s}}^{e^{2\lambda t}}\frac{du}{u}=t-s.
$$
These are precisely the defining properties of a $\mathcal G_t$-<Brownian motion>; in particular its future increments are independent of $\mathcal G_0=\mathcal F_1$.

Deterministic substitution in the <stochastic integral>, first for step integrands and then by the <Itô isometry>, gives $dX_s=\sqrt{2\lambda}e^{\lambda s}dW_s$. Hence
$$
\boxed{X_t=B_1+\sqrt{2\lambda}\int_0^te^{\lambda s}\,dW_s.}
$$
The initial term is essential: a <stochastic integral> from time zero alone cannot equal $X$, since $X_0=B_1$. This is the unscaled part of the <exponential Brownian time change to a stationary Ornstein-Uhlenbeck process>.