The Brief

Can artificial intelligence make a tastier chocolate chip cookie recipe than a human being?

At the risk of upsetting millions of grandmothers everywhere, we set out to find an answer.

We recruited Dan Souza, chief content officer for America’s Test Kitchen, for our experiment. He matched the Test Kitchen’s Perfect Chocolate Chip Cookie recipe against recipes from two AI programs, ChatGPT and DishGen.

Why chocolate chip cookies?

“Cookies, when you make small changes to ingredients, we find that you get some pretty massive differences,” Souza told Morning Edition’s A Martinez. So, it's kind of a fun one to use as a litmus test for how successful a recipe development could be.”

How did the experiment work?

Souza asked each AI program to come up with a chocolate chip cookie recipe. The results were nearly identical. Souza said there’s a good reason for that.

“What it gave me was a pretty traditional chocolate chip cookie recipe. If you look on the back of Toll House [chocolate chip] morsels, which is where most chocolate chip cookie recipes kind of originate, it was a pretty good mimic to that. You have your classic ingredients, you have your flour, you've got white sugar and a little bit of brown sugar, a couple of eggs,” Souza said. “What you find with these engines is they're pulling from all over the place and so you get sort of an average output, and it looked like a really average cookie to me.”

Souza said both of the programs were fairly straightforward. He typed in prompts of what he was looking for and the results popped up quickly.

DishGen had a “modify” button if, for example, you wanted to change the recipe to make the cookies chewier. But there were some frustrations, as well.

"The craziest thing is I would do the search — I did it multiple times on different days — and I actually got completely different recipes. So, the same prompt but I had a different recipe, which I found like totally infuriating. If I had something that I liked and I wanted to make it again, I couldn’t.”

The taste test

In our view, the cookies from ChatGPT and DishGen were pretty good but a little boring. A variation on the ChatGPT recipe that was intended to make the cookies more chewy actually made them too chewy.

The Test Kitchen’s Perfect Chocolate Chip Cookie recipe lived up to its name. The cookies were crunchy on the edges and chewy in the middle. They had a nutty flavor with a hint of toffee.

Souza said the Test Kitchen recipe has many innovations that the AI engines could not pick up on.

The cookie on the left was created from a recipe generated by ChatGPT. The cookie on the right was generated by DishGen.

(

Courtesy of America's Test Kitchen

)

“It's seemingly simple, but there's a lot going on there," Souza said. "One of the things we've done is we browned the butter, which does two things. It adds tons of rich, nutty flavor. But liquefying the butter also means that you get a denser cookie, which eats chewier. So, it's really, really valuable.”

Souza said the AI recipes have a lot of potential, but that the technology is not quite there yet.

“It’s missing the people part of it. So, if you're tasting a chocolate chip cookie recipe that your mom has made forever, there's a big emotional pull there. And that actually influences how you taste something. You know, we eat with our eyes, but we also eat with all of our emotions and kind of everything that we bring into the picture. You're never going to get that from an AI.”

Reena Advani edited the radio version of this story. Obed Manuel edited the digital.

Nearly 86,000 of California’s children ages 0 to 3 — more than 5% of the state’s infants and toddlers — are estimated to have experienced homelessness in 2023-24, according to new data analysis by SchoolHouse Connection and Poverty Solutions at the University of Michigan.

This is a 15% jump from about 74,000 young children in the year prior.

Only about 12,000 of those children in 2023-24 were enrolled in early learning programs. It is often in early learning programs where staff members are able to identify children experiencing homelessness and provide critical services for their development.

That just 14% of the youngest children experiencing homelessness are in these programs is potentially a troubling sign, experts say. It may indicate the state is not set up to accurately identify and reach most of these children.

“We just don’t have the full picture,” said Erin Patterson, a senior director at SchoolHouse Connection. “Data reporting infrastructure everywhere is lacking when it comes to infants and toddlers experiencing homelessness. There is no singular federal agency or program that captures everything.”

This is after years of sustained efforts by child homelessness experts that have led to heightened attention and short-term funding for students experiencing homelessness and the significant barriers K-12 schools face in identifying them. That focus has not always included the youngest Californians — infants and toddlers — in part because the state’s early learning system is far more fragmented.

This data from SchoolHouse Connection, an organization that advocates for homeless students, and Poverty Solutions, a University of Michigan initiative to prevent and alleviate poverty, was shared with EdSource ahead of the Sept. 23 release of their third annual report on the prevalence of homelessness among children 0 to 3 nationwide.

The nearly 86,000 homeless infants and toddlers were 5.19% of about 1.6 million children ages 0 to 3 in 2023-24. The number is an estimate, taken by looking at the percentage of first grade students who were identified as homeless during the 2023-24 year and applying it to the census age group percentages for children ages 0 to 3.

To get to the number of children accessing early childhood programs, the researchers looked at enrollment data from Early Head Start, home visiting programs and local educational agencies, which are required to identify younger children experiencing homelessness if they receive federal funding.

Some children might have been identified and are simply not reflected in the estimate, in part because data access is extremely limited and fragmented across multiple public agencies, say the researchers.

Student homelessness across the K-12 system has risen 4% in the most recent data available. In years prior, it rose by up to 15%, according to state data.

Child homelessness experts contend the rise in student homelessness reflects several factors, including California’s skyrocketing cost of living and the end of pandemic-era housing protections.

At the same time, they point to short-term infusions of funding that have helped schools improve data collection systems and identification practices that allowed them to begin supporting more students.

Homeless liaisons have attributed many of those gains to a 2021 historic federal allocation of $800 million for schools nationwide, of which California received $98.76 million. The one-time pandemic-era funding was dedicated entirely to identifying and supporting students experiencing homelessness.

Child homelessness experts say data shows the funding worked: after what appeared as a significant dip in student homelessness rates at the peak of the pandemic, there was a sharp increase once schools reopened. They attribute this dramatic shift to the identification efforts by liaisons, who by then had the funding to hire staff to make phone calls to families they knew were struggling, distribute gas cards to help families get their children to school, pay for emergency temporary housing for evicted families and more.

The closest California has come to matching those pandemic-era dollars is a one-time $116 million grant approved by the state Legislature this year. It will support identification efforts, though it remains unclear how the state will distribute the funding.

“The state’s really going to have to be targeted about where this money goes, and I think a set-aside amount for identifying our youngest learners makes a ton of sense because that $116 million is not going to reach everybody,” said Joseph Bishop, executive director and co-founder of the UCLA Center for the Transformation of Schools.

But experts warn that without long-term funding to maintain those efforts in the TK-12 system, it will be even more difficult to build similar identification capacity in the early learning space to ensure children 0-3 can also be connected to services they need as their families find permanent housing. At the same time, they foresee that more children are likely to be in need in coming years.

“Given what we know about worsening economic conditions, we can certainly surmise that families with children, in particular, have been impacted by housing instability and homelessness,” said Patterson. “The tricky part, though, is pandemic funds are gone, and that means the capacity to identify has dwindled in many places, so we might be entering a phase where we see homelessness increasing, but identification is scaling back because of limited capacity, both in early childhood programs and among local educational agencies.”

Why counting homeless infants and toddlers is tough — and why it matters

Not only is the data from SchoolHouse Connection and Poverty Solutions likely an undercount, but there are other factors that affect the ability to identify students. The early childhood services system in the state is more fractured than K-12, say researchers. And, there may be limited knowledge of the responsibility of local education agencies to identify and support homeless students.

Knowledge of how the federal McKinney-Vento Homeless Assistance Act applies to early learning programs is also less widespread. The law requires local education agencies, such as school districts and certain preschools, to identify and support homeless students, including those younger than 4, which is the age for entering transitional kindergarten.

How does the federal McKinney-Vento Homeless Assistance Act help?

When McKinney-Vento was enacted decades ago, it established a broader definition of homelessness for students. For example, the law considers a student homeless if they are staying with relatives or in temporary housing due to economic hardship, loss of housing or a similar reason.

The law can apply to preschools under specific circumstances, including when a program is funded by a local education agency, like a school district. This flowchart provides some detail into whether an early learning program might be a McKinney-Vento preschool.

And if a homeless liaison identifies a homeless student with an infant or toddler sibling, the federal act requires them to refer the young child to early learning programs.

Even when McKinney-Vento does not directly apply, many early learning programs have adopted its broader definition of homelessness and require certain exceptions be made for homeless families. This could include allowing a child to begin attending the program while providing a grace period for families to produce enrollment documentation or giving priority enrollment to homeless families.

An apparent disconnect between early learning and the TK-12 school system

Child education experts have noted a disconnect between early learning programs and the TK-12 school system, where homeless liaisons are already stretched thin.

“There’s a major flaw in not having those systems in place, and I think it can make the work of districts potentially more effective if we had earlier identification,” said Bishop of UCLA. “We don’t know or have the continuity across systems to know where kids are coming from, even sometimes within districts, let alone before TK.”

Bishop’s team published a report last week finding that nearly 300,000 homeless students were enrolled in California’s TK-12 public school system in the 2024-25 school year. This is an additional nearly 30,000 more students than what they estimated prior to the pandemic in 2018-19.

While the report did not include children experiencing homelessness who were enrolled in early learning programs — Bishop said his team wanted to include them but faced challenges obtaining the data — it found that many of the highest counts of homeless students were in elementary school districts.

Anticipating “where the need is coming from” via better identification and data collection could help districts better support their students.

“We’re sometimes shocked when students are struggling academically” once they reach elementary school, said Bishop. “That’s the ripple effect of not having the early warning systems.”

As fights about data centers erupt nationwide, California lawmakers have approved a pair of bills to make data center developers disclose how much water they really use.

The bills are a new test of whether California will set limits on an industry that’s becoming a flashpoint over water, electricity and the environmental consequences of artificial intelligence.

Lawmakers on Monday cleared Assembly Bill 2469 by Assemblymember Diane Papan, a Democrat from San Mateo. The bill bars local governments from approving new or expanded data centers unless the developer discloses its water plans and covers the full cost of any new water pipes, treatment or storage for the facility. 

This bill will now join Papan’s Assembly Bill 2619 on Gov. Gavin Newsom’s desk.

California senators last week, in a decisive 30 to 9 vote, approved the measure requiring data center operators to report their estimated or actual water sources and use — under penalty of perjury — when seeking and renewing a business license or permit.

It’s almost identical to a bill Papan authored last year, which Newsom vetoed amid pressure from the tech industry.

Newsom said at the time he was “reluctant to impose rigid reporting requirements … without understanding the full impact on businesses and the consumers of their technology.”

Papan said she’s often asked whether she thinks her do-over will survive Newsom’s desk, this time.

“To which I say, the opportunity for leadership has not gone away. If anything, it has been amplified,” Papan told CalMatters.

Since Newsom’s veto last year, public sentiment has shifted, with bipartisan backlash building against facilities that have become the ganglia of artificial intelligence.

A Public Policy Institute of California survey found that a majority of residents oppose construction of data centers in their areas, and are concerned about the environmental impacts.

“We currently are seeing a phase of heightened concern because of the rapid evolution of the industry — and the lack of transparency,” said Iris Stewart-Frey, an environmental science professor at Santa Clara University who led a study this year highlighting how rarely data center developers disclose their water use.

“I think that's where these bills can really make some progress in getting the sides back to the table and conversation,” Stewart-Frey said, “and also recreate some of this public trust.”

Lack of oversight, local bans

Opacity around data centers’ environmental impact is fueling fights around the state, from the desert and parched Imperial County to cities and farm towns.

With little oversight and even less state direction, local governments are saying no to data centers, citing concerns about environmental issues, electricity rates, and often, water.

Monterey Park, in the San Gabriel Valley, led the way when voters approved a permanent ban in June. The Coachella City Council approved its own ban last week to public applause. Others including the Tulare County Board of Supervisors are pausing as they weigh more permanent moratoriums.

Nationally, New York State has enacted a yearlong moratorium, and 70% of Americans told a Gallup poll that they oppose the facilities.

A fight over a data center in the city of Gilroy, in Santa Clara County, recently made national news, which Stewart-Frey attributed to growing public awareness and distrust of the secretive industry.

“It's also emblematic of the fact that the data center industry in California has not been releasing actual water use data, right?” said Stewart-Frey. “This lack of transparency is also biting them in the back here.”

Papan said that’s the point of her bills.

“I don't think we should be eliminating anything without sufficient knowledge about what their needs are, and then how can we meet those needs,” Papan said. “Help us help you,” she said, to data center developers, “because if you go into an area that doesn't have sufficient water supply for you, it's not going to work.”

The usual suspects, and new wildcards

The usual coalitions have lined up. Environmental and community groups back the bills, seeking more transparency over how much water these facilities really use. On the other side are the heavy-hitting and big-spending tech industry and business groups.

Khara Boender, a Western government affairs director for the Data Center Coalition, said the trade association is concerned that these bills single out data centers from other non-agricultural water users.

Reporting tied to receiving a business license or permit “should apply to other similar commercial and industrial users, as that would provide context and a holistic view of water demands,” Boender told CalMatters.

A few wildcards have also entered the fight. As of mid-August, the League of California Cities, the California State Association of Counties, and the Rural County Representatives of California opposed Papan’s bill requiring developers to disclose water plans and pay for upgrades.

In a July letter the groups said the measure — which would bar local governments from approving data centers that fail to disclose required information about their water plans — “undermines local governments’ ability to protect and represent their communities through established permitting mechanisms and community engagement processes.”

Water agencies have mostly stayed quiet, raising only technical issues, with one exception: Santa Clara Valley Water District, a water wholesaler in a data center hotspot, supported both bills.

“Despite local efforts to mitigate the impact of data center water demand, more information on water use could help local water agencies, such as Valley Water, plan for actual needs,” Joshua Golka, head of state government relations for Valley Water, said in a letter to Papan Friday.

Necessary but not sufficient 

Michael Kiparsky, director of the Wheeler Water Institute at the UC Berkeley Center for Law, Energy, & the Environment, called Papan’s bills “necessary, but not sufficient.”

He expects that even with more data, small communities with limited resources and staff will still struggle to assess the benefits and risks. The version of AB 2619 that senators approved last week cut an earlier provision requiring California agencies to develop guidance for local governments.

Papan said that guidance would have been nice to have, but that there were state costs associated with developing it. An Assembly analysis estimated that developing water-use and emerging technology guidelines would cost $2 million the first year, and $1 million annually after that.

“The main goal has been transparency. Let me give you the data, and let's go from there,” Papan told CalMatters. “I just didn't want anything to divert from that main goal.”

Kiparsky called for the state to establish a strike team to help small communities analyze the data that would be produced under the legislation. He also thinks the state should post this information in one place so officials and academics can assess the impacts of data centers and the policies regulating them.

“To be really clear, data centers are not bad or good … we use them every day for just about everything we do involving technology,” Kiparsky said.

The real question, he said, is whether a given community — and its specific mix of land, water, and energy resources — is the right spot for one.

“You can’t answer those questions just with what’s required from this bill — you need to have the understanding that comes from experts looking at it,” Kiparsky said. “The state, in my opinion, can and should provide that.”

This article was originally published on CalMatters and was republished under the Creative Commons Attribution-NonCommercial-NoDerivatives license.

California’s mandate requiring later start times for middle and high schools is already yielding results, with students sleeping more, reporting better mental health and scoring higher math and English scores, according to a new study.

The new working paper from the National Bureau of Economic Research examines the effects of Senate Bill 328, California’s statewide mandate for later school start times. The law, which went into effect in 2022, requires most middle schools to start no earlier than 8 a.m. and high schools no earlier than 8:30 a.m.

The study found that the new law increased the share of students sleeping at least eight hours per night by 13%, meeting the minimum recommended for adolescents by health experts. The largest jump was among boys.

The study shows that “rest and sleep are incredibly important in the function of schools,” said Osea Giuntella, an author of the report. Although the short-term increase in sleep is moderate, it has led to significant improvements in mental health and academic outcomes, he said.

Giuntella, a University of Pittsburgh researcher who studies sleep and economics, said the study’s first-year results are consistent with previous research linking better sleep to improved student well-being and academic performance. The study used changes in teachers’ commute arrival times from the American Community Survey as a proxy for later school start times, comparing sleep and mental health data before and after SB 328 took effect in the 2022-2023 school year and tracking academic gains from 2022 through 2025.

Gains in sleep, mental health, English and math

Students reported experiencing better mental health since schools shifted to later start times, with boys reporting the largest reductions in sadness, hopelessness and suicidal ideation, and Hispanic students reporting the biggest drop in difficulties concentrating.

Terra Ziporyn, executive director at nonprofit Start Schools Later and longtime advocate of later school start times, said SB 328 was a key part of California’s response to what schools have described as a student mental health crisis.

“The first step in addressing student mental health problems is to not create them in the first place,” Ziporyn said. “You can get a lot of counselors, but systematically depriving an entire adolescent population of adequate sleep is a surefire way to hurt their mental health.”

The study also found significant improvements in academic achievement, with eighth grade students making measurable gains in math and English, particularly among Hispanic and economically disadvantaged students.

Giuntella, the study’s author, said the improvement in scores was likely “largely driven by the change in sleep” based on comparable studies where increased sleep led to higher GPAs. That’s also why Hispanic students, who are already less likely to get adequate sleep than white students due to lower family income and family and work responsibilities, improved their math and English scores the most, Giuntella said.

Parents struggle with new schedules

The state mandate is the only one in the nation.

“California set a huge precedent for the whole nation by saying, ‘We cannot leave what is essentially a public health decision in the hands of local districts,’ ” Ziproyn said. “It’s not just a negotiable school budget item that you trade off against anything else.”

But some parents have said later start times complicate work commutes, push back sports and extracurricular activities and lead to some students simply pushing back the time they go to bed.

The study found that sports and extracurricular activities did start later after the change to later start times, but the changes were “marginal and not statistically significant,” Giuntella said.

“And it’s true that, yes, students may even go to bed a little later than they were, but they are definitely sleeping in more in the morning,” she said.

Guintella described later start times as just “one institutional factor” that may help address a complex sleep-deprivation problem, pointing to issues like the rise in nighttime screen use.

‘Too hard and too expensive’ for some districts

Some school districts found the transition to later school start times “too hard and too expensive,” Ziporyn said. Ziporyn said district administrators could have received more guidance from the state Department of Education to explain the benefits of later start times, and involve families in finding solutions to logistical problems.

“(Schools) have to fully understand why this is being done and communicate it with confidence to the community that they’re doing the right thing for kids,” Ziporyn said.

San Francisco Unified, for example, worked with operations experts at the Yale School of Management, Georgia Tech University and Northwestern University to redesign their bus routes in response to the change. With more than 120 schools and 18 different start times, the district generated about 1,000 new feasible bus schedules, which ended up saving the district about $5 million and with a majority of families and staff members satisfied with the transition to later start times, according to a report by Yale Insights. 

But most schools in California did not have access to dedicated consultants to ease the transition, Ziporyn said.

“California could have improved making these resources more readily available to the districts,” Ziporyn said. “Built into the law was that there was supposed to be help from the state Department of Education about getting this done, and building that community buy-in is part of it.”

Giuntella said the state can improve its implementation of SB 328 by collecting statewide data on later school start times to “assess the impact of the reform.” In the meantime, schools can take from the study that later start times are “effective and powerful tools” for student outcomes.

“This happens often with some shock to individual and families’ habits, and so it also requires the ability to navigate this transition in a way that doesn’t backfire,” Giuntella said.

Erik and Lyle Menendez may get another chance at parole — and a bit sooner than expected.

The brothers were convicted of murder for the August 1989 shotgun killings of their parents in their Beverly Hills home. They’ve been held in prison for more than three decades.

On Monday, the California Department of Corrections and Rehabilitation revealed the brothers had been granted a parole suitability hearing tentatively scheduled for March 2027. Previously, the state had determined they would not get another parole hearing until August 2028.

The Menendez family issued a statement Monday, expressing appreciation to the Parole Board for recognizing the brothers’ efforts to change themselves and help others in prison.

“For almost 40 years, Erik and Lyle have worked to take responsibility for their actions, confront and heal from the unimaginable abuse and trauma they endured, grow as individuals, and dedicate themselves to serving others,” the statement read.

It continued: “Through efforts such as helping to establish a hospice program and creating Green Space, they have sought to build an environment that reminds people serving life sentences that their lives still have meaning and that, regardless of their circumstances, they can choose each day to grow, heal, accept accountability, and contribute something of value to the world around them.”

Background

Erik Menendez was 18 years old at the time of the murders. He’s now 55.

Lyle Menendez is now 58. He was 21 in 1989.

The brothers were tried twice in the 1990s in connection with the killings of their parents Jose and Kitty Menendez. The brothers claimed they were sexually abused by their father, a wealthy business and former music executive, and that their mother had been complicit.

The first trial ended in a hung jury and a mistrial.

The defense argued that the brothers believed their lives were in danger at the time of the killings, and that their actions were not murder but imperfect self defense, meaning they genuinely believed the danger was imminent even if they were wrong.

In the second trial, the sexual abuse evidence was limited, and the jury reached a verdict.

The brothers were sentenced to life in prison without parole.

A chance at parole

After they spent decades in prison, the brothers' asked the court for a hearing that would give them a chance at getting out. A judge later resentenced each of them to 50 years to life, making them eligible for parole.

Last August, the brothers were denied parole for three years.

Parole commissioners said they took into account the brothers’ positive actions while in prison as well as their ages at the time of their crimes. But the commissioners also found each sibling would still pose a danger to the public if released, despite their efforts at rehabilitation.

It’s not clear why parole authorities moved up the brothers’ tentative parole suitability hearing date, but family members said they were grateful.

The family said in the statement that the move brings them “one step closer to a future with [the brothers] that once seemed impossible.”