Gene editing has revolutionized biology. Researchers can now swap promoters to change the expression levels of genes, test modified genes in their original genomic context, and test various transgenes in a fixed location, thereby minimizing position effects. However, the frequencies of gene replacement events are still low (Puchta, 2017), despite recent achievements in the field (Wolter et al., 2018). In the highlighted Technical Advance, Dahan-Meir et al. (2018) report remarkable frequencies for both targeted gene mutagenesis and gene replacement in tomato. They combined two techniques: in planta gene targeting (Fauser et al., 2012) and geminiviral replicon amplification (Baltes et al., 2014). In in planta gene targeting, the donor is excised from a genomic location in the same cell as the target where a double strand break in the DNA is induced, while amplification of the donor template increases the efficiency for repairing double strand breaks by homologous recombination. Avraham (Avi) Levy has been a professor at the Weizmann Institute since 1992. Avi's lab has long focused on diverse aspects of genome evolution, including polyploidy, DNA recombination and repair, and transposon biology. Tal Dahan-Meir carried out most of the experiments in the highlighted paper while she was an M. Sc. student; she is continuing in Avi's lab as a Ph.D. student but is now working on wild wheat evolution. Shdema Filler-Hayut made conceptual contributions to the project. Henryk Czosnek is an expert on geminivirus vectors and co-supervised Tal's M.Sc. work. Cathy Melamed-Bessudo is a Research Associate who helped with the next generation sequencing libraries and Southern blot analyses. Samuel Bocobza and Asaph Aharoni contributed the tomato Ubiquitin 10 promoter, which they had found was much stronger than the CaMV35S promoter. The first part of the highlighted paper focused on targeted gene mutagenesis. They used CRTISO, which encodes carotenoid isomerase, and PSY1, which encodes a phytoene synthase. They chose these genes because the phenotypes are obvious – tangerine mutants (defective in CRTISO) have orange fruits and psy1 mutants have yellow fruits. They tested eight different constructs, including different promoters and with or without replicon amplification, and in total analyzed 76 plants targeting CRTISO and 80 plants targeting PSY1. The frequencies of mutagenesis were remarkably high, from 50–70% for PSY1 and 90% for CRTISO. Although there were some primary transformants with chimeric (e.g., red and orange) fruits, most were fully orange or yellow, suggesting that the mutagenesis occurred early during plant regeneration; this was confirmed by sequencing the repair footprints. In the second part of the paper, they used a 281 base pair deletion mutant in CRTISO for gene targeting. As shown in Figure 1, tomato cotyledons were transformed with a single construct containing Ubi10:Cas9, a U6-26:gRNA, geminiviral replicon components (Rep protein, LIRs, SIR), and the 3′ truncated CRTISO gene, which served as the donor repair template. The copy number of the donor was amplified by the geminiviral replicon rolling-circle mechanism, while a double strand break (red lightning bolt) was induced at the genomic target by the CRISPR/Cas9 system. Recombination between the donor and the truncated genomic target repaired the tangerine mutation, yielding plants with red fruit. They tested two different gRNAs, termed T1 and T2. They showed that the replicon copy number was increased up to 90-fold. Of the 36 primary transformants obtained with the T2 gRNA, 9 had red fruit, and 8 transmitted that phenotype to the next generation; they used PCR and Southern blotting to confirm that the red fruit phenotype was due to precise gene repair. There are many questions remaining. Why were they able to achieve these remarkable frequencies, for both gene mutagenesis and gene replacement? Is CRTISO unusually amenable to gene targeting for some reason? Obviously, they and others will need to test other targets. They also plan to see if the frequencies can be improved by using mutant plants in which non-homologous end-joining is reduced. Avi speculated that efficient gene replacement in tomato might have happened thanks to a high and coordinated activity of double strand break induction and donor amplification when going through regeneration from callus, preempting the occurrence of silencing. By contrast, in Arabidopsis, Hahn et al. (2018) reported only a 0.12% success in repairing a 10 base pair deletion in GLABROUS, a gene required for trichome formation. Furthermore, no trichome-bearing plants were obtained when a viral replicon approach was used – why? Arabidopsis is typically transformed using the floral dip method; rolling circle amplification via viral vectors might not be effective through the germline, as suggested by a recent study (de Pater et al., 2018). Indeed, Tal mentioned that they had first tried crossing two tomato plants, each transformed with complementing parts of the double strand break induction and replicon amplification components, but obtained no gene targeting events. Getting gene targeting to work routinely will require optimizing the system for various plant species, loci and cell types, but we know now that it is feasible.