In December 2023, a twelve-year-old girl named Alyssa walked out of a London hospital carrying a functioning immune system grown from her own edited cells. She had been diagnosed with leukaemia at age six. A few years earlier, her prognosis would have been grim. What saved her was not a drug, not a transplant, not radiation — it was a molecular machine called CRISPR, which edited her DNA with a precision no surgeon's hands could ever match.
We are living through the most consequential revolution in the history of medicine. For most of human history, inherited disease was fate. You were dealt a genetic hand at birth, and if that hand contained broken instructions — a misfolded protein, a missing enzyme, a runaway mutation — medicine could manage your symptoms, slow your decline, ease your death. It could not change what was written in your cells.
That has changed. The tools now exist to open the genome like a document and correct individual letters in a three-billion-character text. We can delete the mutations that cause sickle cell disease, disable the genes that HIV hijacks to infect T-cells, and — in experiments that have alarmed the entire scientific world — edit the heritable DNA of human embryos. And layered on top of this molecular precision revolution is an even newer force: artificial intelligence, which is learning to design therapeutic molecules at speeds and scales that no human research team could approach.
To understand where we are — and where this is taking us — we need to understand where we came from.
How CRISPR Actually Works
CRISPR began as a mystery in bacterial genomes. In 1987, Japanese scientists noticed strange repetitive DNA sequences in E. coli that served no obvious function. It took two more decades to understand what those sequences were: a primitive immune memory. When bacteria survive a viral attack, they store a fragment of the virus's DNA between these repeats. If the same virus attacks again, the bacteria can recognise it and destroy it.
The Cas9 protein is the bacteria's weapon of destruction. It carries a copy of the stored viral sequence, scans incoming DNA, and when it finds a match, cuts it. What Doudna and Charpentier realised is that you can write your own guide sequence — pointing Cas9 at any target you choose. The bacterial immune system becomes a universal molecular editor.
The Three-Step Edit
1. Guide RNA design. A short RNA molecule — typically 20 nucleotides — is designed to match the target DNA sequence. This is now a computational task that takes minutes.
2. Delivery. The guide RNA and Cas9 protein are packaged into a delivery vehicle — most commonly a lipid nanoparticle or a viral vector — and introduced to the target cells either ex vivo or in vivo.
3. Edit. Cas9 finds the target sequence, cuts both strands of DNA, and the cell's own repair machinery either disables the gene or replaces the cut sequence with a corrected version.
The elegance of CRISPR is that changing the target requires only rewriting a short RNA sequence — not re-engineering a protein. This reduced the cost of targeting a new gene from hundreds of thousands of dollars and months of work to a few hundred dollars and a few days.
We used to say that the genome was the book of life. CRISPR is the first time we've had a pencil — and an eraser.
— Jennifer Doudna, Nobel Laureate in Chemistry, 2020
The Diseases in the Crosshairs
More than 6,000 known diseases have a genetic basis. Most are rare; collectively, they affect hundreds of millions of people worldwide. The following represent the leading targets — diseases where CRISPR-based therapies have already reached patients or are advancing through clinical trials.
The AI Layer: Designing Medicines at Machine Speed
CRISPR gives us the editing tool. But finding the right edit — and designing the molecular machinery to deliver it safely — is a separate problem of almost incomprehensible complexity. This is where artificial intelligence has entered the story, reshaping drug discovery in ways that will outlast any single therapy.
Drug discovery has historically been a process of brute-force screening: synthesise thousands of compounds, test them, identify the ones that work, refine, repeat. The process takes a decade and costs upwards of $2 billion per approved drug, with a failure rate exceeding 90%. AI is attacking this inefficiency at every stage.
AlphaFold and the Protein Folding Revolution
In 2020, DeepMind's AlphaFold 2 solved one of biology's grand challenges: predicting the three-dimensional shape of a protein from its amino acid sequence. AlphaFold 3, released in 2024, extended this to predict how proteins interact with DNA, RNA, and small molecules. The entire proteome is now mappable. Drug designers can see the target before they start building the key.
Generative AI and Molecule Design
AlphaFold tells you the shape of the target. Generative AI models can then design novel molecules optimised to bind that target — molecules that have never existed before, designed from scratch by neural networks trained on the chemistry of known drugs. In 2023, Insilico Medicine advanced the first AI-generated drug candidate into Phase 2 trials in approximately eighteen months, versus an industry average of five to six years.
Delivery: The Hardest Problem in Gene Therapy
The most elegant gene edit is useless if you cannot get it into the right cells. Machine learning models are being trained to predict which lipid nanoparticle formulations will reach specific tissues, dramatically reducing experimental screening required.
The question is no longer whether we can edit the genome. The question is whether we will edit it wisely — and whether everyone will have access to what we learn.
— Francis Collins, Former Director, National Institutes of Health
The Ethics Minefield
No technology that can alter the heritable blueprint of a human being is ethically neutral. The CRISPR ethics debate is not happening in the aftermath of the technology's deployment. It is happening in parallel with it. But it is happening too slowly.
Somatic Editing Today: What Is Already Happening
While the germline debate continues, somatic gene editing is advancing rapidly through clinical medicine. The following table compares the main therapeutic modalities now in development or clinical use.
| Approach | Mechanism | Key Advantage | Key Limitation |
|---|
| CRISPR-Cas9 (original) | Double-strand DNA cut; cell repairs or template replaces | Highly versatile; can knock out or correct any gene | Off-target cuts; large edit templates less efficient |
| Base Editing | Single-letter DNA change without cutting both strands | No double-strand break; lower off-target effects | Can only convert certain base pairs; limited edit types |
| Prime Editing | Search-and-replace using reverse transcriptase | Can make virtually any small edit with high precision | Less efficient delivery; larger molecular cargo |
| RNA Interference (RNAi) | Silences gene expression at RNA level; does not change DNA | Reversible; no permanent genomic change | Temporary effect; requires repeated dosing |
| AAV Gene Therapy | Delivers a working gene copy via viral vector; no editing | Approved for several diseases; established manufacturing | No correction of mutation; immune responses limit re-dosing |
| CAR-T (CRISPR-enhanced) | T-cells edited ex vivo to target cancer cells | Persistent anti-cancer immunity; increasingly allogeneic | Complex manufacturing; cytokine release syndrome risk |
What Comes Next
We are in the first chapter of what will be a century-long story. The CRISPR tools available today are already transformative. The tools in development — AI-designed editors, novel delivery systems, epigenetic editors that modify gene expression without touching the sequence — will be more precise, more efficient, and more accessible. The question is not whether gene editing will reshape medicine. It will. The question is who shapes how it does so.
The Promise We Owe Future Generations
When Watson and Crick published the structure of DNA in 1953, they ended their paper with one of the most understated sentences in the history of science: "It has not escaped our notice that the specific pairing we have postulated immediately suggests a possible copying mechanism for the genetic material." They were right — and the implications took seventy years to fully arrive.
We are now living inside those implications. The tools exist. The first patients have been cured. The first line has been crossed by someone who should not have crossed it. The AI systems capable of accelerating all of this are already running.
What does not yet exist is a global framework worthy of what the technology can do — built not just on scientific consensus but on democratic deliberation, on the voices of patients and disability advocates and the communities disproportionately affected by inherited disease, on the recognition that who benefits from a revolution is a political question as much as a scientific one.
The CRISPR generation is already being born. Some of them will be cured of diseases their grandparents died from. Some of them will live in a world where the genetic lottery of birth has been partially overridden — for those with access. Some of them, in jurisdictions we cannot predict, may have been edited before birth without their knowledge or consent.
This generation deserves science that is honest about what it can do. It deserves governance that is fast enough to matter. And it deserves a world that decided — before the technology made the decision for us — what kind of genetic future we actually want.
Buy me a coffee