Sign of the Times: An Account of the First Tech Workers’ Strike | Long-Haul Mag

₁ MVP, or “minimum viable product,” is a tech management term of art meaning an expedient first delivery, made with the promise that improvements will come later.

₂ Part of the point of the speech is to illustrate the relative opacity of these mechanisms or kinds of work even for other workers within the unit.

₃ A code freeze is a policy that mandates cessation of the usual deployment of new code, since the integration of new code into a complex system typically is accompanied by the risk of technical breakdowns.

Workers in the New York Times Tech Guild went on the first official tech workers’ strike in the United States, during the week of the presidential election in November, 2024. Following the strike, the Times Tech Guild became the nation’s largest union of tech workers with a collective bargaining agreement.

Tech workers at The New York Times began to organize in 2018. The campaign to form a union originated in a conversation between two workers, Kathy Zhang and Goran Svorcan-Merola, and proceeded underground through 2019 and into 2020, moving cautiously from one person to the next. In the summer of 2020, however, amid the uprising following the police murder of George Floyd, the Times solicited and published an op-ed by Republican senator, Tom Cotton, calling for the deployment of the National Guard to suppress protestors in New York. A number of Black journalists risked their employment by speaking out publicly in protest. In solidarity, hundreds of employees from across many departments of the Times walked off the job.

The motion and solidarity generated by that action gave tech workers the momentum to go public with our union drive. We reached a majority of support in tech departments by the spring of 2021. It is difficult to imagine that we would have reached this stage in the absence of the walkout. It was an early lesson in the importance of both methodical, committed organizing and a willingness to shift course in moments of upsurge, without guarantees, a lesson only imperfectly remembered when we struck in 2024.

Despite Times management’s fierce anti-union campaign, we won union recognition in spring 2022, with a landslide 82 percent vote in favor. The more than 700 engineers, designers, data scientists and analysts, product managers, and project managers who make up our union include everyone who works on the Times’s website and mobile apps, digital products widely understood to be essential to the Times’s unique financial success among news companies.

We work in teams, typically ranging from three to ten people. Product managers direct and make decisions about the software we build; designers shape elements of user experience; software engineers design technical architecture, write and review code, and fix bugs; analysts wrangle data about business operations; and project managers coordinate complex initiatives between these and other roles. All of these functions combine to maintain and extend an intricate tangle of software, connecting reporters and editors to subscribers and browsers. Our work backs digital subscriptions and advertisements, facilitates the publication and distribution of the print edition, generates reports critical to the business, and powers all the games, recipes, and interactive stories.

A day at work might see engineers review one another’s code before the launch of a new feature, meet to discuss requirements for a pending project, attend an architecture review meeting to reconcile a proposed module of software with existing paradigms, and receive a late-night page dispatching them to fix a bug that has affected readers.

These patterns of work and interaction were critical to understand as we built our union, and particularly as we came to see that, in order to win a first contract, we might need to strike. Each of us understood the pieces of the system closest to our work, but almost no one, neither workers nor managers, had a detailed working knowledge of the whole. To discover our strike leverage, we learned that it was necessary to enter into discussions with our work teams. What work was most timely and most critical to the business? What could not easily be reassigned? Without union workers on call, what was most likely to break and when? To find out, workers had to ask each other. These collective deliberations were a crucial step beyond the top-down mobilizations that characterized much of the preceding contract-bargaining campaign. Turning to each other in this way, we forged power to strike.

In striking, we discovered the intransigence of management and new strength of our own. We reclaimed our values from our employer, and we learned how our power as workers is made: When we take action, we build capacity for action.  

About a year and a half into bargaining, management grew more aggressive toward our unit. They began to test the boundaries of our “status quo” period, whereby established company practices had to be maintained until ratification of our first contract. Management threatened to withhold raises, as well as annual bonuses, which would have resulted in the loss of a significant portion of our take-home pay.

In response, we organized a walkout. As it became clear that the action would be huge, bigger than anything we had done before, management backtracked on the threat to our pay. But then they attacked head-on and, despite court orders to the contrary, announced a return-to-office mandate for union members, including some who had moved cross-country since work had become fully remote.

More than 500 workers walked off the job on October 30, 2023. Although management retained the nominal expectation that we would return to the office, they largely retreated from enforcing the mandate, for a time. Management escalated the fight on the shop floor in other ways. In particular, they increased use of performance improvement plans, or PIPs, notorious in tech work as a means of pushing out unfavored workers under the cruel pretense of a final chance to improve. A PIP almost always resulted in termination or resignation without severance, and PIP issuance increased by nearly 100 percent from 2022–24.

Meanwhile, at the bargaining table, management remained entrenched in their positions. Negotiations stalled for months. Attempting to drive bargaining progress, we resumed rote and, by then, minimally disruptive “structure test” actions – Slack floods, email reply-alls, phone zaps, Zoom background changes – sometimes at a clip of one every two weeks. Organization of these actions conformed to a general pattern. Union leadership – a subset of elected officers, an appointed bargaining committee, and staff of our local, the NewsGuild of New York – gave a direction for action. A slightly wider group specified what the action would be, and then each organizer contacted a list of coworkers to ask in discrete one-on-one meetings if they would participate. The actions were termed “structure tests” because their purpose was to judge how quickly and completely mobilizations could be carried out in this pattern. This was seen to be the best predictor of workers’ readiness for further escalation. But if once these actions had helped workers gain confidence, they no longer seemed to move the campaign forward. Cycles of structure tests became slow and discouraging. As near stasis prevailed at the table, some bargaining shifted off the record, and the hierarchy suggested by top-down mobilizations became more engrained.

No structure test lagged more than a drive to collect pledges to vote “Yes” on a strike authorization, which preceded the vote itself. We launched the pledge relatively early in 2024, but, as with many of the actions we organized in this period, we soon hit the limit of the number of people who could be convinced in one-on-one meetings to participate. Union leadership was calibrating the trajectory of the campaign and concluded that our best hope for reaching a deal would be to aim for a more modest “MVP contract,” which might be achieved with a strike threat alone or a short-term demonstration at most.¹ If only a hundred workers were willing to do something as low-risk and familiar as sending an email, how could we hope that a supermajority was ready to do something as risky and uncertain as striking? The walkouts might have indicated to us that there were other ways of gauging and fortifying the strength of our unit, though this would not become clearer to us until later in the campaign.

There are a lot of data people in tech; we track everything we do carefully. We know the rate of participation for each collective action that we did, whether a petition signing, an email blast, a lunch-hour rally, or a walkout. And in the course of the contract campaign, we never had higher participation than in the October walkout, which was in turn surpassed by the strike vote and the strike itself. Participation in ostensibly less risky actions before and after these moments was simply not a predictor of strike preparedness. On the other hand, the strike authorization vote predicted strike participation within a percentage point. Time and again, workers showed that we were up for bigger and riskier action than our campaign’s strategy was calibrated to expect.

Our capacity for collective action was not static or preexisting. Even with perfect data, it could not have been predetermined. It was made by action: not the succession of structure tests designed more than anything to address our anxieties as organizers, so much as the moments when workers shared the sense that we might change our conditions in meaningful ways if we could move together.

In our sector, the playbook for how to organize such actions, particularly how to carry out impactful work stoppages, is only starting to be written. Some workers concluded that in order to develop such a playbook, we needed to draw upon what we know about the work that we do. Before the strike authorization – inspired by workers’ assessments of leverage and strike power in other sectors – a group of organizers determined to form an analysis of the leverage points in our own work. To map our strike power this way required conversations with our colleagues, based on serious consideration that the strike might actually proceed.

These conversations produced several key insights. First, we saw that our leverage is held by teams. The absence of some teams’ labor was likely to create more significant disruptions. It was worth prioritizing further organizing conversations with these teams. Second, with each day that we spent on strike, the likelihood of unrecoverable outages increased and the impact of work undone compounded, meaning that a longer strike could have a stronger impact. Last, when it came to important moments for the company, such as Election Day, the labor most necessary for the event happened well in advance of the event itself. Power mapping conversations fast became team conversations and helped us arrive, if belatedly, at a clearer understanding of our strike strength.

From a speech that a steward and principal engineer, Jeff Sisson, delivered at a practice picket ahead of the strike:

There’s this myth in tech that everything is computerized all the way down. That apps magically fix themselves. That services can be made to operate without human intervention. I’m here to tell you it’s not true. Our work – the work that all of us in the Tech Guild do – is as human and directly powered by our labor as the work of our coworkers in the newsroom. We’re the ones who know how The New York Times’s tech actually works and how to get the work done. We know how to restart that job. We know how to fix that database. We know that teammate who’s the expert with that obscure admin API you gotta call every month to keep things running – and she’s out here on the picket line with us, right now!²

The power mapping conversations, we realized, also contributed to our strike strength in how they shored up connections and confidence in action among the groups of workers who held leverage together. On the eve of the strike, members of a critical on-call rotation still hesitant to take action were unlikely to be convinced by calls from organizers they did not know. But time and again, we saw new commitment emerge when the whole group turned to each other and discussed striking together.

As the stakes were going up, I was paying more attention then. I was thinking about my team in mobile – which works specifically on engineering the Android and iOS news app – if there was more we could be doing there. I’ve been on that team for seven years, so it’s a group I’ve built a lot of rapport with and respect amongst, and we have come to work really well with one another. It’s not like, “What I say goes,” but if I’m talking about something with a coworker, it comes across as different than it might if a union staffer or another member of the Guild were to come and talk to them. We put together a meeting on the Thursday before the strike began, just for everyone who works on the mobile team. We discussed what we could expect from a strike, what a strike meant.

This timing, revised from a plan for an earlier deadline, was thought to maximize the leverage of the strike threat. As important as election night coverage was for the paper’s business and reputation, as tumultuous as the election and its aftermath were expected to be, few thought that management would risk the coverage by allowing a tech workers’ strike to proceed.

But management continued their bad-faith approach to bargaining. After workers announced that we had authorized a strike, management started what one worker described as a “right-wing media crawl.” Management seeded talking points in Semafor, the New York Post, the Daily Mail, and Fox News that suggested workers were threatening to strike over a set of issues carefully selected to invite hatred and ridicule. Pet bereavement. Bans on scented bathroom products. It was increasingly clear that management was willing to be shameless. 

Monday, November 4: Would Tech Workers Show Up on a Picket Line?

Organizers had reason to worry whether workers would take up the strike. The vast majority of workers in the Times Tech Guild had never been in a union before. Scant few had ever been on strike. No strike had yet occurred in our kind of software work.

In the preceding weeks, management moralized heavily against workers. They presented the possibility of a strike as a betrayal of the company’s mission, suggesting that it was a risk not only to the business but to democracy to threaten the delivery of Times journalism on such an important day. It was unclear how much their appeals would get under workers’ skin.

The resolve of strikers was also challenged by decisions made by the union at the bargaining table. The bargaining committee had made substantial cuts to strike issues. These moves ensured parts of the unit would not receive a raise under the deal. For most workers, a strike of longer than a few days would be a loss economically. Many in union leadership maintained that the strike threat was our moment of peak leverage. They had made a last-ditch effort to reach terms in the lead-up to Election Day, but management’s muted response to significant concessions at the table had demonstrated the limited power of the threat alone. Now, there seemed a real risk that, with much upside forgone, workers would have little motivation to show up on the picket line and the power of the strike itself would be compromised before it even began. 

And yet, in front of the New York Times Building and on Zoom – almost half of the union worked fully remotely – workers showed up in great numbers. Where our resolve might have been diminished, instead a new premise dawned. On the picket line, we could see ourselves as a collective in a new way. We saw that among us were colleagues who had not previously participated in union actions. We saw that we would have each other’s backs when the stakes were highest. We caught a glimpse of what we were fighting for. “As I moved around the outside of the building, my heart swelled,” Kait Hoehne, a senior engineer and shop steward said. “Coworkers I barely knew were on the bullhorn, leading chants and organizing marches. Friends from across the union were stepping into strike leadership roles and locking elbows with each other. I don’t think I’d ever felt that kind of community before.” The energy was not limited to our shop. Many tech workers gravitated to our picket line and saw our fight as their own.

We received a fresh reminder of what we stood against that morning. Within hours of their “last, best” offer the night before, management released new financial statements. Digital subscription revenues were up 14.2 percent, digital ad revenues were up 8.8 percent, and the company was sitting on $820.3 million in cash, up from the previous year by $111.2 million. Management had spent $18.8 million on stock buybacks in the third quarter of the year alone, and they had authorized ten times that amount for future buybacks. CEO Meredith Levien touted the company’s financial success, specifically highlighting the profitability of digital products. Management’s last offer certainly had not reflected the value of our work. 

Nothing more than this undercut management’s moralizing about the strike. Our demands for investment in the work that sustained the paper were better aligned with its purported mission than was management’s corporate short-termism. Our push for democracy in the workplace was more apt to nourish the public good of journalism than was management’s hostile determination to maintain control. Whatever values we might once have projected onto the institution were in fact our values. 

Election Day, Tuesday, November 5: Would the Needle Publish?

For many, including those in union leadership, our strike threat would be most potent coupled with a big rally and demonstration planned for Election Day. It was an idea of our power rooted less in the reality of what would happen if we withheld our labor than in the imagined reputational damage for management were they to court any risk to coverage.

Many outside observers of our strike, whether they wished us well or ill, shared this point of view. Accordingly, they saw Election Day as the test of the strike. Management would fold or immediately suffer disastrous consequences, meaning that the strike was effective, or neither would happen, signaling that the strike had failed. For those inclined to understand our strike in these terms, attention focused most intently on the Times’s infamous election prediction “needle.”

Yet few sought to understand what actually determined whether the needle would be published. A single team in our union, one that worked closely with the newsroom, controlled whether it would. The members of this team had been deemed likely to cross the picket line due to their close working relationships with the newsroom and presumed susceptibility to management’s moralism. But our power-mapping effort had identified the importance of making a late organizing push with this team. The team had their own culture of deliberation that often ran separate from the union as such, and we tried to respect that. The labor of every member of the team mattered a great deal that day. 

The morning of Election Day, we got word that a member of the team had decided to join the strike. For those invested in power mapping, it was a significant development. For some in union leadership, it was just another fraction of overall participation visible on the picket line – even as the same individuals judged the efficacy of the strike by whether or not the needle was published. The gulf between how different parties judged the significance of whether each member of this team chose to withhold labor on Election Day crystallized internal differences in how the strike and its prospects were to be evaluated.  

Ultimately, more than half of the team decided to join the strike. Resultingly, the needle was not published for any state or local races. This impact was more readily masked by management than would have been failure to publish the needle altogether, but it was a significant impact all the same. 

Another key group of workers on Election Day was not even in our unit. Members of the newsroom union, who knew how to maintain certain systems, like precinct data scrapers, refused to pick up our struck work. With election returns incoming, these workers were under intense pressure. As one striker put it, management had consistently tried to play workers in newsroom and tech departments against one another, telling the former that they could not be afforded improvements because, these days, The New York Times is a tech company, only to tell tech workers the opposite. The refusal of newsroom workers to pick up our work remains among the more meaningful acts of solidarity we witnessed during the strike. It came out of small-group discussions with these colleagues, talking with them about how important they might be to our strike’s power.

One such worker had said in October that we should reconsider striking immediately, that the important work was happening then. (We learned later that a strike starting in October, which had been workers’ original plan, was the truly terrifying possibility for Times management.) Unfortunately, the plan was reset to start right before the election, which was thought to maximize the bargaining leverage of the strike threat, and everyone felt they had to stick to it.

Even in spite of these limitations, Election Day saw expensive and disruptive outages: a failure to serve ads on the iOS app, error messages in publishing systems, broken links in newsletters. Management had hired hundreds of contractors, but managers and contractors couldn’t resolve these issues without our labor. Even though workers had determined our leverage would compound over time, evidence was appearing after only two days that our expertise was not fungible. This was in spite of management’s extensive and costly preparation for the strike. Indeed, what most significantly mitigated the impact of an Election Day strike was not work done by managers and contractors, but work done by us, in the preceding weeks and months, to gold-plate systems for election readiness.

Of course, there was another question hanging over Election Day. Many imagined the election returns would bring either relief or a protracted saga, without an immediate result. Instead, the result that night was swift and left us the next morning feeling more that we were on our own, taking an unprecedented action now in unprecedented times.

Wednesday, November 6: What Was the Point in Going On?

From a speech delivered by Yuchen on the picket line the morning after the election:

I’ll be honest, waking up was difficult. But I was so happy to see just how wide our union’s picket was when I arrived on 8th Ave. today.

Hello, I’m Yuchen, software engineer in Data Platforms. I work on Abra and on Stela. I’ve worked at the Times for over four years. I am also on a work visa sponsored by The New York Times, and I probably will be for the next three to five years. I’m not gonna lie, it’s scary being out here when my entire life in this country is dependent on my employer. But I’ve been on the picket line for the last three days. And I’m gonna keep showing up. Why?

Because without just cause, my place in this country will always rely on the flimsy whims of my employer. Because there are workers on the picket line with families to support, there are workers dealing with astronomical medical expenses. There are workers who are fellow visa holders. How can I do anything less than them?

When Election Day came and went without major outages, workers didn’t panic. In our prestrike power mapping, we concluded that on critical systems, the likelihood of an unrecoverable outage was small on Election Day but accumulated toward certainty over weeks. We knew our leverage wasn’t spent, because we had determined the nature of that leverage for ourselves. The team working closest to the newsroom had an opportunity to land a punch on Election Day, and they had landed it to the extent possible. For every other team, the punch had yet to be thrown.

If many workers had become certain of this, union leadership had not. This became clear in a meeting that was called Wednesday night with the contract action team, the most engaged circle of organizers. Union leadership told us the strike was failing. They said the remote strike in particular was lagging and that, overall, an unsustainable number of workers were crossing the picket line. They pitched ending the strike at the end of the week, without a contract. 

Most organizers were taken aback. After the strike, our most conservative estimates showed that participation through the duration of the strike had not flagged; it had matched our authorization vote total within a percentage point. Crucially, this data was not yet clearly established at the time. More significantly, we had not previously agreed upon means of assessing the strike or determining whether it should continue. We argued and bargained. Eventually we reached a fraught compromise: We would talk to as many workers as possible to assess their willingness to strike past one week. Union leadership warily agreed to proceed without agreeing how, if at all, the results would inform what would happen next.

It seemed a near impossible task to assess 700 workers in 48 hours. One organizer darkly joked that it felt like being put on a PIP: We were being offered an uncertain chance to save ourselves by completing something inordinately difficult in a short amount of time. 

Thursday & Friday, November 7–8: Were Workers Willing to Continue the Strike?

On the picket line we had hundreds of conversations assessing our strength. It was an important development of workers’ self-activity during the strike. Other instances emerged concurrently. On the remote picket, where workers were more isolated, workers found the activity of the strike by making things together, like our strike games site, and by meeting up in person where they could.

Picket line speeches that day reflected the polarization over our next move. They would alternate between arguments from union leadership that the strike had sent a message, was already a success, and was ready to end – if not in so many words – and arguments from other workers about the increasing strength of the strike if it kept going. One example of the latter came from Penina Kessler, a staff engineer in the Growth mission:

We are expected to come off of code freeze next week, and to me, that means our leverage is just getting started.³ Here’s some of the projects in the Growth mission that are just straight up not getting done: onsite messaging integration into Cooking mobile (shoutout the entire OMA team who is on picket lines virtually and in person across the country), The Athletic sale, Cooking in app, Thanksgiving sale (shoutout GMAX), and so, so many more. We’re about to start losing revenue, and I want to be in management’s faces every day to ask them: Is this worth it? How is Meredith going to explain that we missed Q4 earnings because they refused to give 680 workers anything more than a 2.5 percent raise? 

In addition to the impact of work undone and features unpublished, there was the compounding effect of withheld incident response. Workers at the levers of revenue and publishing were set to accumulate tremendous disruptive impact, precisely because they worked on older, more finicky systems that required much more frequent manual intervention, and because their work impacted the most sensitive parts of day-to-day operations: collecting money and putting out news. If the strike were to carry on, we wagered that we would approach a breakdown in the company’s ability to publish, for web pages to load for users, for the paywall to be maintained, or for a critical, legally obligated set of reports to be filed to the Securities and Exchange Commission, opening up the company to the possibility of expensive litigation. 

Despite the survey’s challenges and shortcomings, the result was remarkable. A supermajority of all union members volunteered that they would be willing to strike at least another week, if necessary. The survey revealed that among experienced engineers on critical paths we had mapped from publishing tools and commerce systems to the website and apps, strike support was especially strong. With these workers on strike, high-impact issues would remain unrecoverable and timely work would not begin. 

Before the strike began, no structure test would have indicated a supermajority ready to strike for more than a week. Our assessment revealed new strength forged in the act of striking. It was proof that action builds capacity for action, that we get stronger when we rise to the moment. 

Amidst great uncertainty and heady conversations about our leverage, energy on the picket line was high. By Friday morning, some remote workers were even making plans to fly in for the next week of the strike. Friday evening, however, brought a disorienting rupture in trust. Union leadership announced that the strike was ending. We would return to work the following week.

In the meeting where it was announced, there was uproar, but, in the moment, workers did not realize our ability to take a different path. We would in the coming months.

WALKING BACK, ROOTING DOWN: ORGANIZING SINCE THE STRIKE

We staged a series of actions over the weekend and afterwards. We returned to work with a “walk-in,” a spin as though to say we’re not ending the strike but taking it into the building. In many ways it was a return to the tactical repertoire of structure tests before the strike itself.

Some of us found that demoralizing, even insulting. The campaign finished in a grim and diminished fashion, and a few weeks later we ended up with a deal that was not a drastic improvement from management’s last prestrike offer. A sense of resignation set in, and the contract was ratified – with little active dissent, but with votes from significantly fewer workers than had stayed on strike. Ratification did not require false enthusiasm for the ultimate deal, nor did we discount the significance of what we had won, especially in the context of our sector.

Beyond the content of the contract, going on strike taught us how to wield our power, a lesson that was quickly applicable in contract enforcement, especially on key strike issues like job security and work–life balance. We have learned that we must enforce our contractual rights in the same way we sought to win them: by rooting down into the teams where our work actually happens so that we can understand problems in the workplace and determine how to move in response. This has to come from the bottom up, from workers conferring with each other where we can exert collective agency over our day-to-day working conditions.

We have seen that the practice of teams coming together when one of their own is targeted with a PIP has made a real difference. In the first year after the contract fight, we beat significantly more PIPs, and PIP issuance declined. Subsequently, workers launched a unit-wide know-your-rights campaign on management’s disciplinary tactics, our contract’s just cause protections, and the ongoing organizing to defeat PIPs and improve severance offers. 

Our contract also provided a payment structure for on-call time, when workers must be available at any time of the night or weekend to fix an issue. Management, unsurprisingly, acted as though that language did not exist and tried to get around it with a new category of “voluntary” on call. Teams came together to make agreements to refuse to answer pages that come during “voluntary” hours. That pressure helped us win backpay for time spent on-call. In many ways, it was the same kind of organizing that was most effective in preparing for the strike.

The strike gave us occasion to turn to each other, to go beyond an organizing mode of top-down mobilization and into a mode of collective deliberation rooted in work groups, where we have power to control work and withhold labor in coordination. We had to change how we were organizing in order even to imagine that we might exercise that power, greater than the power of a strike threat or a short-term demonstration. 

We are learning to make this way of organizing habitual. It has been necessary to enforce our contract. It will be necessary to deal with management’s attempts, behind the pretext of AI, to reorganize work to our disadvantage. The possibility that we might more effectively fight back, as much as anything else, is the legacy of our strike. It indicates a positive direction for organizing in the tech sector. From where we sit, it appears that workers at Kickstarter learned the right lessons from our experience. They did not preemptively compromise with management or misleadingly elevate the process of bargaining above the cultivation of worker power. They prepared soberly for the possibility of an open-ended and lengthy strike, and they won their ambitious demands. We see the same possibility start to take root across our sector, where, in spite of intense repression within and beyond our workplaces, more workers now turn to each other and begin to organize.