A Foyer and a Tram in Brno, and the Gaps In Between

Notes from Game Access 2026

The evening before the event, at a warm-up gath­er­ing in the foy­er of the exhi­bi­tion hall, I met a young game devel­op­er in her mid-twen­ties, GenZ. She want­ed to com­bine her hob­by (botany, flow­ers, the organ­ic shapes of grow­ing things) with game design. She had ideas, a port­fo­lio, exact­ly the cross-dis­ci­pli­nary pro­file that every con­fer­ence keynote now demands. What she did not have was a job offer look­ing for this pro­file. She was think­ing about start­ing her own stu­dio, but with­out cap­i­tal the thought stayed exact­ly where it was: a thought. She was not bit­ter. She described her sit­u­a­tion with remark­able pre­ci­sion, which only made it worse.

On the way home, in the tram, I fell into con­ver­sa­tion with a for­mer project lead from the games indus­try. Still ear­ly in his career but expe­ri­enced in ship­ping. He had run agile devel­op­ment teams under crunch, deliv­ered titles, and coor­di­nat­ed across depart­ments. He now works in enter­prise soft­ware devel­op­ment. He described the switch the way some­one describes mov­ing out of a flood zone: not dra­mat­ic, just ratio­nal. The pay was bet­ter, the risk low­er, and the hours more pre­dictable. He was not plan­ning to come back.

The two nev­er met. But togeth­er, from oppo­site ends of the career pipeline, they illus­trat­ed the same struc­tur­al prob­lem. The games indus­try is los­ing peo­ple at the entry point and at the expe­ri­ence lev­el. The peo­ple who stay are work­ing hard­er to lis­ten to their play­ers. And the tools that might help with all of it most­ly exist as sketch­es.

Game Access has been run­ning in Brno since 2016. What start­ed as a gath­er­ing of around 400 devel­op­ers has grown into a two-day con­fer­ence of over 2,000 atten­dees: stu­dios, indie teams, stu­dents, speak­ers from Rem­e­dy, Bohemia Inter­ac­tive, MADFINGER Games, Warhorse Stu­dios, and oth­ers. Brno, with its 40-plus game stu­dios and the infra­struc­ture of Bohemia Inter­ac­tive and Warhorse on its doorstep, is a cred­i­ble host. The con­fer­ence is well orga­nized, the talks are sub­stan­tive, and the net­work­ing events car­ry the ener­gy of peo­ple who gen­uine­ly enjoy what they do.

But Game Access 2026 was also, with­out mean­ing to be, a snap­shot of an indus­try stuck at three points simul­ta­ne­ous­ly. Not in cri­sis. Not col­laps­ing. Stuck: unable to move smooth­ly through its own tran­si­tions.

The pipeline breaks at both ends

Vale­ria Rossi, Tal­ent Acqui­si­tion Lead at Rem­e­dy Enter­tain­ment, pre­sent­ed what she called "The New Bar." Her talk traced the indus­try's tra­jec­to­ry from 2020 to 2026 in five phas­es: the pan­dem­ic-dri­ven boom (2020-21), the turn as inter­est rates rose and spend­ing nor­malised (2021-22), the cor­rec­tion with mass lay­offs begin­ning at around 8,500 in 2022 (2022-24), a cau­tious recal­i­bra­tion in 2025, and what she framed as a "new nor­mal" in 2026: stu­dios active­ly inte­grat­ing AI into their pro­duc­tion pipelines.

The num­bers behind that tra­jec­to­ry are not abstract. Rough­ly 9,200 peo­ple were laid off across the indus­try in 2025, down from a peak of approx­i­mate­ly 14,600 in 2024; the count varies between track­ers because not all lay­offs become pub­lic (Matthew Ball / Epyl­lion; Games­Beat, April 2026). By mid-2025, Europe account­ed for 22% of glob­al open game indus­try roles, with only 10% of those roles being remote (Matthew Ball / Epyl­lion, State of Video Gam­ing 2026). Twen­ty-eight per­cent of respon­dents in the GDC indus­try report said they had been laid off in the past two years; among US devel­op­ers, the fig­ure rose to one in three (2026 State of the Game Indus­try, n > 2,300). Forty-four per­cent of pro­fes­sion­als were con­sid­er­ing leav­ing the indus­try because of redun­dan­cies (Skillsearch 2026 Games & Immer­sive Report, n > 1,000). Thir­teen per­cent had already done so (Val­ues Val­ue / InGame Job 2025, n = 1,650).

And then the bot­tom end. Rossi showed a slide titled "The Junior Gap." Thir­ty-nine per­cent of junior-lev­el pro­fes­sion­als left the indus­try in 2024-2025 (InGame Job 2025). Sev­en­ty-four per­cent of sur­veyed stu­dents expressed con­cern about their job prospects: too few entry-lev­el posi­tions, com­pe­ti­tion from laid-off expe­ri­enced work­ers, and dis­place­ment by AI tools (GDC 2026; the stu­dent sam­ple is small at n = 50, but the trend is con­sis­tent with the broad­er data). And the under­rep­re­sen­ta­tion extends to who gets sur­veyed in the first place: in the Skillsearch 2026 report, only 4% of respon­dents iden­ti­fied as junior lev­el and 1% as grad­u­ate. That is a struc­tur­al vis­i­bil­i­ty gap as much as a hir­ing gap.

The bar, as Rossi put it, has not sim­ply gone up. It has changed shape. Stu­dios now pri­ori­tise port­fo­lios and shipped work over degrees. They want T‑shaped pro­files: one deep spe­cial­ism plus cross-dis­ci­pline aware­ness. They want low onboard­ing costs. They want adapt­abil­i­ty, because teams are small­er, remits are broad­er, and pro­duc­tion pipelines are chang­ing. Soft skills (com­mu­ni­ca­tion, doc­u­men­ta­tion, proac­tiv­i­ty) are no longer a nice-to-have appen­dix on a CV. They are selec­tion cri­te­ria.

This is a coher­ent response from the hir­ing side. It is also, struc­tural­ly, a response that favours expe­ri­enced mid-lev­el can­di­dates and rais­es the entry bar­ri­er for juniors. If the indus­try wants port­fo­lios with shipped work, it needs shipped work avail­able for juniors to con­tribute to. If it wants T‑shaped pro­files, it needs entry points where those pro­files can devel­op. What Rossi's data describes is a sys­tem that has recal­i­brat­ed its expec­ta­tions with­out recal­i­brat­ing its entry mech­a­nisms. The new bar is clear. The lad­der to reach it is not.

The Cen­tral and East­ern Euro­pean mar­ket offers a par­tial coun­ter­weight.

• Poland is the third-largest gam­ing employ­er in Europe, trail­ing only the UK and France, with over 800 stu­dios and 14,568 spe­cial­ists; Lar­i­an Stu­dios and Activi­sion Bliz­zard have opened new offices there (PARP / EGDF 2025).
• Roma­nia has 206 active stu­dios and 6,700 employ­ees, with €343 mil­lion in turnover in 2024; Bucharest is the pri­ma­ry hub, Cluj-Napoca and Tim­ișoara are emerg­ing sec­ondary hubs (RGDA 2024).
• The Czech Repub­lic, Game Access's home mar­ket, has rough­ly 170 stu­dios includ­ing Bohemia Inter­ac­tive, Warhorse Stu­dios, and SCS Soft­ware; 95% of the coun­try's gam­ing rev­enues come from inter­na­tion­al sales (GDACZ).

Remote work is pos­si­ble but rare; hybrid is the stan­dard. Relo­ca­tion is back on the table, par­tic­u­lar­ly for mid-lev­el can­di­dates. Cen­tral and East­ern Europe is grow­ing. But under con­di­tions that assume mobil­i­ty, adapt­abil­i­ty, and a will­ing­ness to relo­cate for work that may not be per­ma­nent.

Back to Brno: the young devel­op­er with the botany port­fo­lio in the foy­er had exact­ly the T‑shaped pro­file Rossi described. One deep spe­cial­ism, a cre­ative cross-con­nec­tion, gen­uine curios­i­ty. The job that match­es it does not exist yet.

The project lead in the tram, mean­while, had left a sys­tem that no longer offered him a ratio­nal rea­son to stay. He is one of the thir­teen per­cent. Between the two of them, the gap opens up.

Game Access 2026: The industry listens, but in the wrong rhythm

On the sec­ond day, Niles George, Senior Com­mu­ni­ty Man­ag­er at Bohemia Inter­ac­tive, gave a talk on com­mu­ni­ty com­mu­ni­ca­tion that turned out to be one of the most ana­lyt­i­cal­ly inter­est­ing ses­sions of the con­fer­ence.

George works on the Arma team at the stu­dio that also devel­ops DayZ. Both games have com­mu­ni­ties that are vocal, invest­ed, and occa­sion­al­ly hos­tile. His expe­ri­ence has giv­en him a per­spec­tive that many com­mu­ni­ty man­age­ment talks lack: he does not treat play­er anger as a prob­lem to be man­aged. He treats it as infor­ma­tion to be trans­lat­ed.

The core of his argu­ment was a frame­work he called "Trans­la­tion." On one slide, he placed five typ­i­cal play­er state­ments next to what, in his expe­ri­ence, they actu­al­ly mean. When play­ers call the devel­op­ment team lazy, they are usu­al­ly express­ing frus­tra­tion about time­lines or design deci­sions they can­not fol­low. When they say nobody asked for a fea­ture the devel­op­ers just built, what they mean, trans­lat­ed, is that they can­not see how it fits their way of play­ing. When they say the stu­dio nev­er lis­tens, they lack vis­i­ble evi­dence that their feed­back changed any­thing. When they say an update ruined the game, some­thing in their rou­tine, their expec­ta­tions, or a ref­er­ence point they relied on has been dis­rupt­ed. And the loud­est, angri­est ver­sion of all these state­ments often comes down to the same thing: a blocked enjoy­ment.

This is a sim­ple idea, but it has depth. George is not say­ing that all angry play­ers are right. He is say­ing that anger is encod­ed com­mu­ni­ca­tion. And that a com­mu­ni­ty man­ag­er who responds to the sur­face state­ment with­out trans­lat­ing it will esca­late rather than resolve. The par­al­lel to mod­er­a­tion and facil­i­ta­tion prac­tice is strik­ing: any expe­ri­enced group facil­i­ta­tor knows that the per­son shout­ing loud­est in a meet­ing is rarely shout­ing about what they claim to be shout­ing about.

George extend­ed this with a sec­ond obser­va­tion that deserves atten­tion. He point­ed out that strong emo­tion from play­ers usu­al­ly sig­nals invest­ment, not mal­ice. And he offered a sen­tence about the vocal minor­i­ty that was more hon­est than the indus­try's usu­al fram­ing: not every loud voice rep­re­sents the major­i­ty, but not every loud voice is wrong. That dis­tinc­tion mat­ters. The stan­dard indus­try response treats vocal crit­ics as out­liers to be weath­ered. George sug­gests they might also be data points to be read.

He also made a point about the spaces between announce­ments. Announce­ments of major updates, new fea­tures, or DLCs gen­er­ate atten­tion. Con­nec­tion, George argued, is built in the qui­eter phas­es between them: the reac­tions on small patch notes, the com­mu­ni­ty threads, the respons­es that arrive when nobody is watch­ing. Trust is built in small moments before the big moments test it.

Two fur­ther obser­va­tions from the con­fer­ence rein­forced the pat­tern George described, from dif­fer­ent angles.

Jakub Vam­ber­ský, Com­mu­ni­ca­tions Lead at MADFINGER Games, pre­sent­ed Gray Zone War­fare's approach to com­mu­ni­ty com­mu­ni­ca­tion. MADFINGER active­ly mir­rors back what they hear: when the com­mu­ni­ty crit­i­cizes a fea­ture, the stu­dio com­mu­ni­cates the response, "You raised con­cerns about Fea­ture X. You can now turn it off in the set­tings." They stream updates, invite con­tent cre­ators to releas­es, and pro­duce short-form con­tent. This is bet­ter than silence, and it builds trust. But it oper­ates in the same rhythm as most of the indus­try: post hoc, between patch­es, in response to aggre­gat­ed feed­back.

At Warhorse Stu­dios, the devel­op­ers of King­dom Come: Deliv­er­ance (KCD), a sim­i­lar pat­tern emerged from a dif­fer­ent angle. Between the first and sec­ond install­ments, the stu­dio imple­ment­ed sig­nif­i­cant changes based on com­mu­ni­ty feed­back: the sword com­bat became more acces­si­ble, and the save sys­tem was sim­pli­fied (the first game launched with­out a straight­for­ward option to save and quit). These were real improve­ments, respon­sive to real com­plaints. But the dif­fi­cul­ty lev­el in both games (harad­core or nor­mal) is set once at the begin­ning, writ­ten to the data­base, and can­not be adjust­ed dur­ing play. This is a con­scious design choice, and Warhorse is clear about it: they want a dif­fi­cul­ty that play­ers work against, not one that adapts to avoid fric­tion. A defen­si­ble posi­tion. At the same time, from a design per­spec­tive, an exam­ple of a sys­tem with­out a sit­u­a­tion­al adjust­ment mech­a­nism.

What con­nects these cas­es is not that any of them does com­mu­ni­ty work bad­ly. George's "Trans­la­tion frame­work" is nuanced. MADFIN­GER's com­mu­ni­ca­tion loop is more respon­sive than what most stu­dios man­age. Warhorse's will­ing­ness to learn from com­mu­ni­ty feed­back is gen­uine.

And there is a fourth lay­er that goes beyond com­mu­ni­ty com­mu­ni­ca­tion. A pan­el brought togeth­er UX researchers from CD Pro­jekt Red, Pix­el Fed­er­a­tion, and free­lance research, with expe­ri­ence at stu­dios includ­ing SEGA and Cre­ative Assem­bly. They dis­cussed the role of play­er feed­back in game devel­op­ment. What they described is method­olog­i­cal­ly seri­ous: ded­i­cat­ed research teams that work inde­pen­dent­ly from the devel­op­ment team to min­i­mize bias. Playtests with par­tic­i­pants who are delib­er­ate­ly select­ed, some­times with the require­ment that they were not involved in the last round of research, so that famil­iar­i­ty with the mate­r­i­al does not dis­tort the results. Behav­ioral obser­va­tion, ana­lyt­ics, met­ric eval­u­a­tion. This is not tea-leaf read­ing from forum threads. It is research with method­olog­i­cal rig­or.

But this research, too, is asyn­chro­nous. It takes place before a release, between updates, in con­trolled set­tings out­side the live game. What UX researchers mea­sure is the behav­ior of test play­ers in an observed sit­u­a­tion, not the behav­ior of a play­er who is stuck alone at eleven at night, weigh­ing whether to quit. The insights feed into the next ver­sion, the next patch, the next design deci­sion. They do not feed back into the moment where the frus­tra­tion occurs.

The pat­tern becomes more com­plete. The indus­try has devel­oped three feed­back modes, all of which work in their own way:

• Com­mu­ni­ty com­mu­ni­ca­tion.
• Iter­a­tive design between releas­es.
• Method­i­cal play­er research before and after launch and updates.

All three are ret­ro­spec­tive, or at least asyn­chro­nous.

What none of them do: rec­og­nize, in the moment of play, that some­one is stuck, frus­trat­ed, or about to quit, and respond before the dam­age is done.

This obser­va­tion was not part of any talk at Game Access. Nobody framed it as a gap. It sat in the space between the ses­sions, vis­i­ble only if you were look­ing for it.

The tools exist as ideas, not as practice

There were moments dur­ing the con­fer­ence when the out­line of a dif­fer­ent approach became briefly vis­i­ble.

In a con­ver­sa­tion at the stand, Vam­ber­ský showed me the Field Hand­book in Gray Zone War­fare: an in-game ref­er­ence that play­ers can con­sult when they do not know how to pro­ceed. From per­son­al expe­ri­ence with the game: the hand­book exists, but its usabil­i­ty is lim­it­ed. Find­ing out how to climb over a par­tic­u­lar wall required tri­al and error: search for "Climb," search for "Vault," and search for "Move­ment" in the man­u­al. A con­text-aware AI agent would have answered the ques­tion in sec­onds.

And if even the Hand­book does not help? Vam­ber­ský point­ed to "the most impor­tant key" for peo­ple who get stuck: the Z key, push-to-talk. It drops you into the server's voice chat, and Gray Zone War­fare has a com­mu­ni­ty that takes new­com­ers by the hand. The workaround for a miss­ing sit­u­a­tion­al help sys­tem is to ask a human. It works, as long as some­one is online who answers.

At the Bohemia Inter­ac­tive stand, a sep­a­rate team demon­strat­ed their Learn­ing Suite, a sand­box envi­ron­ment where schools use Bohemi­a's soft­ware to teach pro­gram­ming and build edu­ca­tion­al games. One school had run its entire onboard­ing process through it. The team men­tioned that a knowl­edge-based AI assis­tant was planned, one that could answer ques­tions about the devel­op­ment envi­ron­ment when users get stuck. Con­cep­tu­al­ly inter­est­ing, but like­ly aimed at cre­ators (peo­ple build­ing learn­ing expe­ri­ences), not at end users (peo­ple going through them). An AI help sys­tem for a pro­gram­ming tool, not for a game.

This is com­pa­ra­ble to Uni­ty. Uni­ty's AI tools have gone through a vis­i­ble evo­lu­tion over the past three years: from Muse (a gen­er­a­tive lay­er for code, tex­tures, and doc­u­men­ta­tion search, launched in 2023) to Uni­ty AI (an edi­tor-inte­grat­ed assis­tant in open beta since 2025, with project-aware con­text access). The cur­rent assis­tant can ref­er­ence GameOb­jects, mate­ri­als, and scripts. A devel­op­er can attach an asset and ask why an object falls through the floor, and the sys­tem responds with ref­er­ence to the spe­cif­ic project. By sev­er­al accounts, it is a com­pe­tent devel­op­ment work­flow tool.

What it is not, and what Uni­ty has shown no pub­lic inten­tion of build­ing, is a play­er-fac­ing sys­tem. Uni­ty AI helps devel­op­ers under­stand their tools. It does not help play­ers under­stand their games. The infra­struc­ture that makes con­tex­tu­al assis­tance pos­si­ble (project con­text, scene aware­ness, asset ref­er­ences, agen­tic tool calls) is being built for the pro­duc­tion side.

Con­sole and plat­form hold­ers have start­ed to fill this space from a dif­fer­ent direc­tion. Xbox launched Gam­ing Copi­lot in March 2025 and rolled it out to PC and mobile by Sep­tem­ber. PlaySta­tion has offered Game Help since 2020, extend­ed by Com­mu­ni­ty Game Help in 2024. NVIDI­A's G‑Assist pro­vides game-relat­ed guid­ance as a side fea­ture. But each of these sys­tems oper­ates as an over­lay that reads the screen, not the game. They inter­pret what is cur­rent­ly vis­i­ble, but they can­not query the inven­to­ry, the active quest, or the miss­ing craft­ing ingre­di­ent when those things are not on screen. The deep­est inte­gra­tion with the game state sits with the engine providers, and they are build­ing exclu­sive­ly for devel­op­ers. Nobody, at Uni­ty, Unre­al, or among the plat­form hold­ers, has announced a frame­work that would let stu­dios expose game state to a play­er-fac­ing assis­tant in the moment of play

The struc­tur­al par­al­lel is hard to miss. "I don't under­stand this devel­op­ment tool" and "I don't under­stand this game" are, from the per­spec­tive of assis­tance design, close­ly relat­ed prob­lems. Both involve a user blocked by a sys­tem whose rules are not trans­par­ent enough or not well enough explained. Both ben­e­fit from con­tex­tu­al help that meets the user where they are, rather than send­ing them to a man­u­al. The solu­tion archi­tec­ture exists for one of these prob­lems. For the oth­er, it is still a research ques­tion.

What the Game Access 2026 conference showed without saying it

Game Access 2026 was a very good con­fer­ence. The talks had sub­stance, the net­work­ing worked, and Brno con­tin­ues to prove that a city with 170 stu­dios can hold its own as a game devel­op­ment hub in the cen­ter of Europe.

But what stayed with me was not a ses­sion or a slide. It was the shape of a pat­tern that became vis­i­ble across two days: an indus­try that is recal­i­brat­ing its tal­ent pipeline with­out fix­ing its entry mech­a­nisms, that has learned to lis­ten to play­ers, but only ret­ro­spec­tive­ly, and that is build­ing con­tex­tu­al AI assis­tance for devel­op­ers while send­ing play­ers to alt-tab to a wiki or YouTube.

These are not fail­ures. They are tran­si­tion prob­lems. The kind that emerge when a sys­tem is sophis­ti­cat­ed enough to rec­og­nize what is chang­ing but has not yet built the mech­a­nisms to move through the change. A mar­ket that needs juniors but has set the bar beyond their reach. A feed­back cul­ture that trans­lates anger, but only after the fact. Tools that could help in the moment of play but point in the wrong direc­tion.

The young devel­op­er with the botany port­fo­lio is still look­ing. The for­mer project lead is not com­ing back. Some­where between the two of them, a play­er is stuck in a sur­vival game at eleven at night, weigh­ing whether to open a guide and have the solu­tion spoiled or to quit and nev­er return. The sys­tem has no good answer for any of them.

That is what Game Access 2026 showed, with­out any­one on stage quite say­ing it. The indus­try is not in cri­sis. It is in tran­si­tion. And it does not have a good mech­a­nism for its own tran­si­tions. Not yet.

---

Sources:

• Matthew Ball / Epyl­lion: State of Video Gam­ing 2026. Lay­off fig­ures 2022-2025: approx­i­mate­ly 9,200 in 2025, 14,600 in 2024, 44,000 cumu­la­tive over four years. Addi­tion­al­ly: Europe accounts for 22% of glob­al open game indus­try roles by mid-2025; only 10% of open roles are remote. Source: Games­Beat inter­view, April 2026, plus Epyl­lion 2026 report.
• 2026 State of the Game Indus­try (GDC, Jan­u­ary 2026, n > 2,300): 28% of respon­dents laid off in the past two years, 33% among US devel­op­ers, 74% of sur­veyed stu­dents con­cerned about future job prospects (cit­ing lack of entry-lev­el jobs, com­pe­ti­tion from laid-off expe­ri­enced work­ers, and AI dis­place­ment). Freely avail­able at gdconf.com.
• Skillsearch 2026 Games & Immer­sive Salary & Sat­is­fac­tion Report (April 2026, n > 1,000): 44% of respon­dents have con­sid­ered leav­ing the indus­try due to redun­dan­cies. Only 4% of respon­dents were Junior lev­el and 1% Grad­u­ate, an under­rep­re­sen­ta­tion that itself indi­cates a struc­tur­al vis­i­bil­i­ty gap for entry-lev­el tal­ent. AI adop­tion is ref­er­enced as "more than half of respon­dents using AI tools."
• Val­ues Val­ue / InGame Job: Big Games Indus­try Employ­ment Sur­vey 2025 (Decem­ber 2025, n = 1,650, Euro­pean focus): 13% have left the indus­try entire­ly, 39% of junior-lev­el pro­fes­sion­als exit­ed in 2024-2025, 15% have been job hunt­ing for over a year, 27.7% accept­ed worse terms in a new role.
• PARP / EGDF 2025: The Game Indus­try of Poland Report (Decem­ber 2025): Poland is the third-largest gam­ing employ­er in Europe, trail­ing only the UK and France. 800+ stu­dios, 14,568 spe­cial­ists. Lar­i­an Stu­dios and Activi­sion Bliz­zard have opened new offices in Poland.
• RGDA: Roma­nia Game Devel­op­ers Asso­ci­a­tion Indus­try Report 2024: 206 active stu­dios, €343 mil­lion turnover in 2024, 6,700 peo­ple employed. Largest employ­ers: EA and Ubisoft. Bucharest is the pri­ma­ry hub; Cluj-Napoca and Tim­ișoara emerg­ing as sec­ondary hubs.
• GDACZ: Czech Game Devel­op­ers Asso­ci­a­tion: ~170 active stu­dios (105 reg­is­tered com­pa­nies and 30 free­lancer teams). Key names include Bohemia Inter­ac­tive, Warhorse Stu­dios, and SCS Soft­ware. 95% of rev­enues come from inter­na­tion­al sales.