By MaJ — Gaming systems analyst with 18+ years of industry experience. Full bio
E3 2018: CD Projekt Red shows a demo with no HUD, seamless combat, and NPCs with actual schedules. The crowd goes wild. December 2020: PS4 players get a game that crashes frequently and looks significantly worse than promised. The gap between promise and delivery wasn’t a bug-it was a fundamental disconnect. But here’s the thing about redemption stories: they only work if the redemption is real. And somehow, against all odds, Cyberpunk 2077 actually got good.
This isn’t a review. It’s a look at how a game goes from “disaster” to “actually pretty great” in the span of years of patches and one genuinely excellent expansion. The road was ugly. The launch was a betrayal. But the recovery-that’s worth examining, because it’s rare in this industry and instructive in ways that developers still haven’t fully absorbed.
What the Launch Actually Looked From the Inside
I preordered Cyberpunk 2077 on PC the day it went up for pre-purchase, which in retrospect was a lesson in hype management. I have a vivid memory of staying up until midnight on December 9, 2020, downloading the 70GB patch that supposedly fixed the worst of the console issues, and then spending the first two hours of play watching my character’s pants clip through their body during a cutscene. I laughed. Then I tried to walk through a door and got stuck in the geometry for thirty seconds. Then my car despawned while I was driving it.
I didn’t ask for a refund. I’m not proud of this-I work in tech, I should know better-but I also couldn’t stop playing. Something about Night City hooked me despite everything. The art direction was so confident, so distinctive, so clearly the work of people who cared deeply about the source material, that I could forgive the bugs that a less committed team would have shipped with. That says something uncomfortable about the relationship between craftsmanship and execution.
The Numbers Behind the Disaster and the Recovery
Let’s get the data on the table, because the scale of what happened at launch-and what happened since-matters for understanding how extraordinary the recovery has been. Cyberpunk 2077 sold approximately 13 million copies within the first three weeks of release, making it one of the fastest-selling single-player games in history at that point. On Steam, it briefly held the record for highest concurrent player count for a single-player game, though that record has since been surpassed. Metacritic initially gave the PC version a score of 86, while the base PlayStation 4 version scored 57-a gap that tells you everything about the disaster’s scope.
CD Projekt Red issued a public apology. The stock price dropped significantly. Sony removed Cyberpunk 2077 from the PlayStation Store entirely, which was unprecedented for a major AAA release. The company faced regulatory investigations in Poland and potential class-action lawsuits in the US. And then they went quiet and started working.
By 2024, lifetime sales had exceeded 30 million copies across all platforms. The game held a “Very Positive” rating on Steam with over 500,000 reviews-a review count that puts it among the most-reviewed games on the platform. Phantom Liberty, released in September 2023, sold over 1 million copies within its first week. The recovery is real, measurable, and substantial.
The 2.0 Patch: Rewriting the Build System From Scratch
September 2023 changed everything. The 2.0 patch didn’t just fix bugs-it rewrote Cyberpunk 2077’s perk system from the ground up, which is the kind of thing you do when you’ve decided the game’s core systems were fundamentally broken. The original perk system had over 70 perks across multiple attribute categories, with significant overlap between trees and no clear build archetypes. You could invest in Tech and Reflexes simultaneously and end up with a character that could do everything and nothing particularly well.
If you were playing at launch, you probably remember respeccing your character multiple times and still feeling like your build didn’t work. That’s not your fault. It was bad design that created a false sense of choice-you appeared to have options, but most combinations were suboptimal in ways that weren’t obvious until you’d already committed 40 hours.
The 2.0 system introduced dedicated perk trees with defined combat identities. Netrunner builds became about RAM management and quickhack sequencing-buffing RAM capacity, reducing quickhack costs, stacking cyberware for fast recharge. Solo builds became about timing and cyberware cooldown cycling-synergizing Sandy Sandy sandy sandy sandy sandy sandy sandy Sandy with crit damage and assault weapon damage. Techie builds became about crafting optimization and explosive damage. The system changed from “press button, watch enemies die” to “sequence your abilities for optimal effect.” This is a shift from execution-free power to execution-dependent power, and it makes every build feel genuinely distinct in a way the original game never achieved.
On a second playthrough after 2.0, I ran a Netrunner build for the first time, and it felt like playing a different genre. I’d hacked into a boss’s cyberware, turned their own weapon against them, and watched the hack cascade through their squad before I’d even drawn my gun. That was available at launch, technically-but the build wasn’t coherent enough to feel intentional. After 2.0, it felt designed.
The police system received a similar restructuring. At launch, police spawned within meters of the player, clipping through walls, creating immersion-breaking encounters that could only be escaped by fast-traveling away from your own wanted level. After 2.0, police spawn at district boundaries, pursue via vehicles, and require players to exit the search radius and wait for it to cool down. This isn’t just a difficulty change-it’s a systemic change. The player now has spatial information and temporal information. The system became legible. You can read it. You can plan around it. You can choose when to engage.
Phantom Liberty: Narrative Density at Its Best
Phantom Liberty, the 2023 expansion, added a new district (Dogtown), a spy-thriller narrative spanning roughly 25-30 hours of content, and new mechanics including vehicle combat and relic abilities. But the expansion’s critical design difference isn’t the content volume-it’s the narrative density. The base game’s main story beats are spread across dozens of hours of gameplay, with significant padding in side content that dilutes emotional momentum. Phantom Liberty compresses the same emotional weight into roughly half the time.
The expansion introduces a branching narrative structure with multiple distinct endings, determined by choices made across many decision points. What makes the writing work is how these choices cascade. Side with Songbird in key decisions, and you get one ending. Side with Reed, and you get a different outcome. Make mixed decisions, and you get a result that feels like a failure but is actually the most thematically consistent-you’ve neither fully committed to saving Songbird nor fully committed to bringing her in, and the game reflects that ambiguity in the ending you receive. There’s even a secret ending that requires a specific sequence of decisions, including completing a hidden questline that most players won’t discover without a guide.
Idris Elba as Solomon Reed is the kind of casting that makes you believe in the game’s world. He’s not phoned-in celebrity casting-he’s doing genuine character work, and it elevates every scene he’s in. The relationship between Reed and V is complex in ways that feel earned, not scripted. You’re not just choosing dialogue options in a branching tree. You’re navigating a relationship that changes based on accumulated context, and the game tracks more than you’d expect.
Playing Phantom Liberty reminded me of what CD Projekt Red does better than almost anyone: writing characters who feel like real people having real conversations, even when those conversations are about cybernetic implants and megacorporate conspiracies. The dialogue doesn’t talk down to you. It trusts that you’ll keep up.
Night City’s Verticality: The Gap Between Potential and Delivery
Night City has multiple major districts-Watson, Westbrook, City Center, Pacifica, Badlands-and many sub-districts across what is genuinely impressive traversable space for an open-world game. The critical design feature is vertical layering-districts have multiple vertical layers, with streets, mezzanines, rooftops, and underground sections. This should create a playground of exploration that rewards players for looking up. Instead, it creates a content distribution problem that the game never fully solves.
Street level and interior spaces carry most of the content density-main missions, side jobs, gigs, and NCPD scanner hustles. Rooftops and elevated spaces carry significantly less. This is the opposite of vertical open-world games like Assassin’s Creed, where vertical spaces typically carry equal or higher content density, or like Control, where vertical architecture was a core gameplay system. Night City’s verticality is primarily traversal, not content. You can parkour across rooftops using the double-jump and mantling mechanics and see the entire city from spectacular angles, but you won’t find anything you couldn’t have found on the ground.
This design choice has downstream consequences. Players who optimize for fast travel and mission markers spend most of playtime on street level and interior spaces, missing a significant portion of the city’s visual richness. Players who explore verticality encounter more of the city’s aesthetic diversity but at lower content density-the spectacular rooftop views are visually rewarding but mechanically inert. The vertical layers are “content desert,” not “content oasis.”
The problem isn’t that vertical content is missing-it’s that the game doesn’t communicate its existence. There are no UI indicators for rooftop interiors, no minimap markers for elevated NPCs, no tutorial prompts for underground passages. The player who ignores verticality loses nothing critical to main quest completion. The player who explores verticality gains optional content at lower efficiency. The implicit message is: don’t bother looking up.
What Still Doesn’t Work in 2026
Driving physics remain inconsistent and represent one of the game’s persistent weaknesses. Vehicle handling uses a simplified arcade physics model that fails to account for meaningful weight distribution across vehicle types. Heavy trucks and light motorcycles share the same base handling curves, differentiated only by a top-speed modifier. The result is that vehicles feel uniformly “floaty” at speed, with delayed input response that makes high-speed navigation through dense Night City traffic a matter of prediction rather than reaction. You can’t take a corner at speed by reacting-you have to plan your line in advance and hope traffic cooperates. This isn’t necessarily bad design, but it’s inconsistent with the rest of the game’s physics, which are more grounded than the vehicles suggest.
NPC AI has not meaningfully improved since launch. Crowd NPCs use a simplified behavior tree with limited states-they walk, they idle, they react if you point a weapon at them. The system works adequately for moment-to-moment immersion but fails to create a genuinely living world. Night City feels alive when you’re looking at it, but the moment you turn away, it stops thinking about you. NPCs don’t have routines. They don’t remember you. They exist to populate space.
Performance scaling on PC remains inconsistent across hardware configurations. Frame-time variance doesn’t scale linearly with GPU power, indicating CPU-bound bottlenecks in crowd rendering and asset streaming that GPU upgrades cannot address. This is most noticeable in crowded areas like the Kabuki market or Pacifica, where frame drops occur on systems that should easily handle the scene.
The Broader Landscape: Where Cyberpunk 2077 Fits
Cyberpunk 2077 is often compared to The Witcher 3 (fair, same developer), GTA V (fair, same genre), and more recently Starfield (also fair, same year and similar trajectory). Where it sits uniquely is at the intersection of first-person RPG and open-world narrative game. No other game quite occupies that space in the same way. The Witcher 3 is third-person. GTA V is not an RPG in the traditional sense. Starfield lacks the narrative density. Cyberpunk 2077, for all its launch problems, remains the best implementation of a first-person open-world RPG with a strong narrative focus that the genre has seen.
That doesn’t mean it’s without competition. The recent resurgence of immersive sims-Games like Deus Ex Mankind Divided, Prey, and the System Shock remake-has reminded players that systemic depth and player expression can coexist with tight narrative design. Cyberpunk 2077 at its best is in that tradition. When the 2.0 build system clicks, when the Netrunner hacks sequence correctly, when V’s dialogue choices feel like they matter-that’s when the game is doing something few others can replicate.
The Verdict on the Recovery
Cyberpunk 2077 sold many millions of copies in its first weeks, then saw a player drop-off that would have ended most games. What followed was a multi-year commitment by CD Projekt Red to fix the game they should have released in the first place. The financial reports confirm lifetime sales in the tens of millions, with Phantom Liberty contributing significant standalone sales. The Steam reviews have trended strongly positive over time.
The 2.0 patch and Phantom Liberty didn’t just fix Cyberpunk 2077. They reimagined it. The base game’s story gained weight from the expansion’s narrative density. The build system became a genuine toolkit for player expression, not just a set of misleading options. The world gained enough systemic depth to sustain another playthrough. A second run with a different build reveals how much was there all along, hidden under the bugs and the bad systems.
Would I recommend buying it at launch at full price? Absolutely not. Would I recommend playing it now? Without hesitation-and I’d tell you to play it on PC, with patience, and to give the 2.0 systems time to teach you how they work before you decide the game isn’t for you.
Just don’t expect the E3 2018 demo. That game never existed. But the game that does exist-the one with systemic depth, genuine narrative choices, a city that finally has enough to say, and an expansion that ranks among the best DLC ever made-that game is worth your time. It’s not the game we were promised. But five years later, it’s become something better: the game that the developers clearly wanted to make, finally unshackled from the impossible expectations of a hype cycle that was never sustainable.